ES3003873T3 - N-glycosylation - Google Patents

Abstract

The present invention relates to a mammalian cell comprising a gene encoding a polypeptide of interest, wherein the polypeptide of interest is expressed by comprising one or more post-translational modification patterns. These modifications are useful, for example, in the production of glycoproteins, where antibodies with the modifications have enhanced antibody-dependent cell-mediated cytotoxicity (ADCC). The present invention also relates to methods for producing the glycoproteins and compositions comprising the glycoproteins, and their uses. (Automatic translation with Google Translate, no legal value)

Description

DESCRIPTION

N-glycosylation

Technical field of the invention

The present invention relates to a mammalian cell comprising a gene encoding a polypeptide of interest, wherein the polypeptide of interest is expressed by comprising one or more post-translational modification patterns. These modifications are useful, for example, in the production of antibodies, where antibodies with the modifications have improved antibody-dependent cellular cytotoxicity (ADCC). These modifications are useful, for example, for improving pharmacokinetic properties, i.e., by attaching PEG or HEP chains to proteins. The present invention also relates to methods for producing the glycoproteins and compositions comprising the glycoproteins, as well as to genome engineering, cell culture, protein production, and protein glycosylation and their uses.

Background of the invention

Glycoprotein biology is the fastest-growing class of therapeutic proteins, and most can only be produced recombinantly in mammalian cells with glycosylation capacity similar to that of humans. The Chinese hamster ovary (CHO) cell has acquired a prominent role as a host cell for the recombinant production of therapeutic glycoproteins, primarily because it produces fairly simple N-glycans with branching and capping similar to those produced in some human cells.

In particular, CHOs produce complex heterogeneous N-glycans with bi-, tri-, and tetra-antennary structures with an o6-fucose (Fuc) core, a minor amount of poly-N-acetyllactosamine (poly-LacNAc) mainly on the a1,6 arm of the tetra-antennary structures, and a unique LacNAc coating with a2,3-linked neuraminic acid (NeuAc).

CHO does not generally produce immunogenic protective structures in non-humans and in humans, such as N-glycolylneuraminic acid (NeuGc) or α3Gal, although these have been reported to occur, perhaps as a result of genetic induction. N-glycosylation can vary for different proteins, as well as for different glycosites within individual proteins, and for example, IgG antibodies are produced with truncated N-glycan structures at the conserved Asn297 glycosite (biantenary structures with α6Fuc core, limited LacNAc, and α3Gal protection).

A major concern with CHO is the substantial heterogeneity in N-glycan processing, which can be difficult to control during bioprocessing and can pose problems for bioactivity and biosafety.

Therefore, an important activity in the bioprocessing of therapeutic products is dedicated to glycan analysis and fermentation control to achieve consistency.

In the last two decades, substantial efforts have been devoted to the genetic glycoengineering of CHO cells with the goal of expanding glycosylation capacity, reducing heterogeneity, and specifically improving or altering sialylation.

All of these studies have essentially used random integration of cDNAs encoding glycosyltransferases and have encountered problems with stability, consistency, and predictability of the introduced glycosylation capacity. The main obstacle has been the need to rely on overexpression of glycosyltransferases and competition with endogenously expressed enzymes due to the lack of simple methods to eliminate them in cell lines.

Thus, only a few of these glycoengineered CHO cells have reached the production stage of clinical therapies. However, a successful glycoengineering strategy has emerged following the discovery that IgGs lacking the o6Fuc core at the Asn297 N-glycan exhibit significantly greater antibody-dependent cellular cytotoxicity (ADCC).

Thus, through a tour-de-force using two rounds of homologous recombination, both alleles of the fut8 gene encoding the α-6-fucosyltransferase that controls core fucosylation were eliminated in CHO, and at least one therapeutic IgG produced in CHO lacking the fut8 gene is now in clinical use.

More recently, the fut8 gene was eliminated using precise gene editing with zinc finger nuclease (ZFN) targeting the gene in a fraction of the time and resources.

The emergence of precise gene editing technologies for knockout (KO) and knockin (KI) has opened the door to a completely different level of speed and ease with which stable genetic manipulation of host cell lines to delete and introduce glycosyltransferase genes can be achieved, and this will undoubtedly have an impact on the engineering of mammalian host cell factories for the recombinant production of therapies.

Therefore, there is a need for mammalian cells, and especially CHO cells, that have specific glycosylation patterns with or without sialylation.

GTf genes

Mammalian cells have a large number of glycosyltransferase genes, and more than 200 distinct genes have been identified, and their properties and catalytic functions in glycosylation processes have been partially determined (Ohtsubo and Marth 2006; Lairson, Henrissat et al. 2008; Bennett, Mandel et al. 2012; Schachter 2014). These genes are classified into homologous gene families with related structural folds in the CAZy database (www.cazy.org). The encoded glycosyltransferases catalyze different steps in the biosynthesis of glycosphingolipids, glycoproteins, GPI anchors and proteoglycans (together referred to as glycoconjugates), as well as oligosaccharides present in mammalian cells. There is enormous diversity in the glycan structures of these molecules, and biosynthetic pathways for different types of glycans have evolved (Kornfeld and Kornfeld 1985; Tarp and Clausen 2008; Bennett, Mandel et al. 2012; Schachter 2014), although our understanding of which glycosyltransferase enzyme(s) catalyze a particular linkage in the biosynthesis of the diverse set of glycoconjugates produced in a mammalian cell is not complete. Glycosylation in cells is a non-template-driven process that depends on a number of factors, many of which are unknown, to produce the different glycoconjugates and glycan structures with a high degree of fidelity and differential expression and regulation in cells. These factors may include: the expression of the glycosyltransferase proteins; subcellular topology and retention in the ER-Golgi secretory pathway; the synthesis, transport, and availability of sugar nucleotide donors in the secretory pathway; the availability of acceptor substrates; competing glycosyltransferases; divergence and/or masking of glycosylation pathways that affect the availability of acceptor substrates and/or result in different structures; and the overall growth conditions and nutritional status of the cells.

GTf isoenzymes

Several glycosyltransferase genes have a high degree of sequence similarity and have been classified into subfamilies encoding closely related putative isozymes, which have been shown or predicted to play related or similar roles in glycan biosynthesis in cells (Tsuji, Datta et al. 1996; Amado, Almeida et al. 1999; Narimatsu 2006; Bennett, Mandel et al. 2012). Examples of such subfamilies include polypeptide GalNAc transferases (GalNAc-T1 to 20), α2,3 sialyltransferases (ST3Gal-I to VI), α2,6 sialyltransferases (ST6Gal-I and II), α2,6 sialyltransferases (ST6GalNAc-I to VI), α2,8 sialyltransferases (ST8Sia-I to VI), β4 galactosyltransferases (B4Gal-T1 to 7), β3 galactosyltransferases (B3Gal-T1 to 6), β3 Gl' cNActransferases (B3GnT1 to 5), β6 GlcNAc transferases (C2GnT1-7 and IGnT2A to C), β4 GalNAc transferases (B4GalNAc-T1 to 4), β3 glucuronyltransferase (B3GlcUA-1 to 3), α3/4-fucosyltransferases (FUT1 to 11), O-fucosyltransferases (O-POFUT 1 and 2), O-glucosyltransferases (O-GLUT 1 and 2), p4GlcNAc-transferases (MGAT4A to C), p6GlcNAc-transferases (MGAT5 and 5B), and hyaluronan synthases (HAS1 to 3) (Hansen, Lind-Thomsen et al. 2014) (See TABLE 1 for an overview of genes and proteins, including the nomenclature used).

Functions of the GTf subfamily

These isozyme subfamilies are generally poorly characterized, and the functions of individual isozymes are unclear. In most cases, the function of isoforms is predicted from in vitro enzyme analysis with artificial substrates, and these predictions have often proven erroneous or partially incorrect (Marcos, Pinho et al. 2004). Isozymes may have different or partially overlapping functions, or they may provide partial or complete support in the biosynthesis of glycan structures in cells in the absence of one or more related glycosyltransferases. Therefore, it is not possible to reliably predict how the deficiency of a particular gene in these subfamilies will affect glycosylation pathways and the glycan structures produced in the different glycoconjugates in a cell, and the in vitro state may not reflect the in vivo state. Furthermore, even more isolated glycosyltransferase genes that are not found among such subfamilies may have unknown functions in glycosylation or capabilities for such unknown functions in the absence of other glycosyltransferase genes.

Knowledge from knockout animals - poor predictability of phenotype This has been particularly evident in studies of knockout mice deficient in glycosyltransferase genes that are members of homologous subfamilies, where striking changes or no apparent changes in glycosylation have been observed (Angata, Lee et al. 2006).

For example, mice with targeted deletion of the p4Galt1 gene encoding the p4Gal-T1 isozyme (one of the 7 members of the pGalactosyltransferase family) demonstrated a partial loss of p4galactosylation of the protein, with a switch from type 2 chains (Gal (31 -4GlcNAc (33-R)) to type 1 chains (Gal p1 -3GlcNAc p3-R) (Kotani, Asano et al. 2004). A corresponding sialylation pattern was observed with repression of the dominant α2,6 and increased α2,3 sialylation characteristic of type 1 chains. Phenotypically, the mice survived to term but showed growth retardation and problematic differentiation of epithelia, as well as endocrine insufficiencies (Asano, Furukawa et al. 1997).

The same lack of global phenotype is observed when individual members of the large family of sialyltransferases responsible for glycan sialylation are deleted (Angata, Lee et al. 2006). For example, mice deficient in the sialylation of core O-glycans catalyzed by the ST3Gal-I enzyme are grossly normal but have markedly decreased numbers of CD8+ T cells in peripheral compartments. ST3Gal-I is one of six other members of the ST3Gal sialyltransferase family, and mice deficient in other family members also mostly display no gross phenotypes. An exception is mice lacking ST3Gal-IV, which exhibit a severe bleeding disorder due to the lack of sialylation and exposed galactose residues on von Willebrand factor and, therefore, its clearance from the circulation. Furthermore, these mice exhibit a leukocyte adhesion defect with decreased activation of P- and E-selectin ligands. Similarly, deletion of individual GalNAc-T polypeptides from the GALNT gene family that initiate GalNAc-only O-glycosylation results in very discrete phenotypes that are often difficult to predict (Bennett, Mandel et al. 2012). The same is true for targeted inactivation of several GlcNAc-transferases that catalyze the branching of O-linked glycans. For example, loss of Core2 GlcNAc-T1 causes only a few obvious abnormalities, although an effect on the functions of E-, P-, and L-selectins is observed. Unexpectedly, compensatory elongation of core 1 glycans was found in Core2 GlcNAc-TI knockout mice, rescuing the phenotype in high endothelial venules (Stone, Ismail et al. 2009).

However, in some cases, the deletion leads to the complete loss of a particular glycosylation ability. In the case of core1 O-glycosylation, deletion of C1GALT1 appears to result in the complete loss of the elongated core1 O-glycans (Wang, Ju et al. 2010), demonstrating that this particular glycosylation function can only be carried out by one enzyme. However, detailed analysis of the function of such isolated genes can often prove difficult if the deficiency leads to early embryonic lethality, as is the case with the C1GALT1 deletion. Another similar example is the deletion of GlcCer synthase (UGCG), which results in the elimination of all LacCer-based complex glycosphingolipids and embryonic lethality (Jennemann, Sandhoff et al. 2005). This is also true for N-glycan production. Macroscopic phenotypes of no or limited viability are observed in several knockout animals with complete or near-complete lack of N-linked glycans (Stone, Ismail et al. 2009). The most pronounced is the effect of loss of the UDP-GlcNAc:dolichol phosphate N-acetylglucosamine-1-1phosphate transferase (GPT) essential for the initiation of oligosaccharide precursor N-glycosylation, resulting in a complete lack of N-glycosylation and lack of viability beyond the E5.5 embryonic stage of gestation. Similar severe phenotypes are observed with loss of the Mgat1-/- encoded glycosyltransferase GnT1, which catalyzes the essential step in the conversion of high mannose into complex hybrid structures (Ioffe and Stanley 1994; Schachter 2014). Knockout mice become nonviable by mid-gestation (E9.5-E10.5) with morphogenic abnormalities including defects in neural tube formation and vascularization.

Deletion of other enzymes that function in the branching of N-linked glycans demonstrated murine phenotypes of variable severity. Deletion of GlcNAc-TII, which catalyzes the formation of complex N-glycans downstream of GlcNAc-I, and o-Mannosidases, demonstrate limited viability in mice, although genetic background-dependent differences have been observed (Freeze 2001). Interestingly, an unexpected increase in the level of Lewis structures has been observed through a compensatory increase in GlcNAc-bisecting structures catalyzed by GlcNAc-TII I encoded by Mgat3 (Wang, Schachter et al. 2002). Unlike deletion of Mgat1 and Mgat2, deletion of Mgat3 and Mgat4a encoding GlcNAc-TIVa are viable without obvious phenotypes. The phenotype of Mgat4b KO mice is currently unclear. Several studies have examined the effect of deleting Mgat5, which encodes the GlcNAc-V transferase. Although no obvious phenotype is observed in younger mice, it has become clear that GlcNAc-V-catalyzed tetraantennary glycans are important for the regulation of T cell activation, leukocyte recruitment in inflammation, as well as other signaling pathways. Thus, it is clear that complete deletion of core elements of certain glycans results in severe phenotypes, whereas deletion of the branching and trimming enzymes often results in unpredictable changes in glycan structures and functional consequences.

Crossbreeding of knockout mice has resulted in mice deficient in up to two glycosyltransferase genes. For example, breeding mice deficient in the sialyltransferases ST8Sia-I (GD3 synthase) and p4GlcNAcT (GD2 synthase), both important for glycosphingolipid biosynthesis, resulted in a complete lack of all types of complex glycosphingolipids. These double-mutant mice were still viable but had a high mortality rate due to extreme sensitivity to audiogenic-induced seizures (Kawai, Allende et al. 2001). Other examples of targeted deletion of two glycosyltransferases include the loss of Fut1 and Fut2. These double-knockout animals present with no clear phenotypes. In addition, mice with targeted inactivation of both Fut4 and Fut7 have been generated. These mice again show no obvious phenotype except a complete lack of E, P, and L selectin ligand activity on leukocytes.

Knowledge of knockout cell lines

In contrast to our knowledge of the effects of glycosyltransferase gene deletion in animals and model organisms, there is very little information on the effects of glycosyltransferase gene deletion in mammalian cell lines. Only a few spontaneous mutants of glycosyltransferase genes have been identified in human cell lines. For example, the LS174T-derived colon cancer cell line LSC has a mutation in the chaperone COSMC that leads to a misfolded and nonfunctional core C1 GalT synthase (Ju, Lanneau et al. 2008). COSMC was originally proposed to encode a core1 glycosyltransferase synthase due to its homology to the C1GALT1 gene; however, it is now thought to be a private chaperone for C1GalT required for folding (Wang, Ju et al. 2010). The COSMC gene is also mutated in the human lymphoblastoid cell line Jurkat (Ju, Lanneau et al. 2008; Steentoft, Vakhrushev et al. 2011). For non-human cell lines, information is basically only available from Chinese hamster ovary (CHO) cell lines, although a similar murine cancer cell line with a spontaneous mutation in the COSMC gene has been found. A series of CHO cell lines (Lee cell lines) with a deficiency in a glycosyltransferase gene were originally generated by random mutagenesis followed by lectin selection, and the isolated lectin-resistant mutant clones were then shown to have defined mutations in the glycosyltransferase genes MGAT1 (Led), MGAT5 (Lec4), and B4GALT1 (Lec20) (Patnaik and Stanley 2006). Several mutant CHO lines have also been found to have mutations in non-glycosyltransferase genes that affect glycosylation, such as CMP-NANA synthase (Patnaik and Stanley 2006) and C4-Glc/GlcNAc epimerase (Kingsley, Kozarsky et al. 1986). Most CHO lines with mutations in glycosyltransferase genes have been found to have partial losses of relevant glycosylation activity, although loss of the MGAT5 gene was found to completely abolish tetraantennary N-glycan synthesis (North, Huang et al. 2010). The latter suggests that there is only one glycosyltransferase gene encoding the p6GlcNAc-transferase activity that leads to tetraantennary N-glycan branching in CHO and that the homologous gene MGAT5C does not serve as a functional backup. The MGAT5C gene has been shown to be involved in the branching of O-Man O-glycans, although some overlapping function might be predicted or expected based on the sequence similarities of the two MGAT5 genes.

Deletion of glycosylation genes in cell lines

The limited information on the effects of glycosyltransferase gene deletions in cell lines is partly due to past difficulties in performing deletions in cell lines before the recent advent of precise gene editing technologies (Steentoft, Bennett et al. 2014). Thus, until recently, essentially only one glycosyltransferase gene, FUT8, had been deleted by a targeted approach using two rounds of homologous recombination in a large-scale effort. Conventional genetic disruption by homologous recombination is typically a very laborious process as evidenced by this FUT8 deletion in CHO, as over 100,000 clonal cell lines were screened to identify a few growing FUT8-/- clones (Yamane-Ohnuki, Kinoshita et al. 2004) (US 7214775). With the advent of zinc-finger nuclease (ZFN) gene targeting strategies, gene disruption has become less laborious. This was demonstrated by the deletion of the FUT8 gene in a CHO cell line, where two additional genes unrelated to glycosylation were also effectively targeted (Malphettes, Freyvert et al. 2010).

More recently, CRISPR/Cas9 editing strategy has emerged and this editing strategy was also used to delete FUT8 gene (Ronda, Pedersen et al. 2014). Furthermore, deletion of mgat1 gene was carried out by targeting ZFN in CHO, which rendered the CHO line unable to produce complex-type N-glycans, and all N-glycoproteins were produced only with high-mannose type structures (Sealover, Davis et al. 2013) (US20140349341). Finally, ZFNs have been used to inactivate the Ggta1 gene (US 20130004992), which encodes a p3Gal-T forming the xenoantigenic epitope Gala1 -3Galp1-4GlcNAc that is highly immunogenic in humans because the human GGTA1 gene is a pseudogene and therefore this particular glycosylation structure does not occur in humans. In all cases of targeted deletion of glycosyltransferase genes, the targeted genes had previously been shown to play unique, non-redundant roles, and it could be reasonably expected that the deletion would result in CHO cells lacking the particular glycosylation capacity performed by the encoded enzyme. However, as discussed above, most glycogens exist in close homolog subfamilies, and the cellular function and role in glycosylation of the encoded isozymes of these paralogs are largely unknown and unpredictable. Furthermore, the loss of any glycosylation capacity may provide acceptor substrates for unrelated and unpredictable glycosyltransferases that produce glycan structures different from those predicted (Holmes, Ostrander et al. 1987).

Thus, it is clear that the genetic engineering of glycosylation genes in mammalian and animal cells is prone to substantial uncertainty, and thus, identifying optimal engineering targets for a given protein will require extensive experimental efforts. Some guidance can be found from the results of in vitro determined substrate specificities of recombinantly expressed enzymes, which are most often performed with truncated secreted forms of the enzymes. While these assays provide general information on the donor sugar and simpler acceptor substrates, they often fail to provide intricate acceptor substrate specificities with respect to more complex acceptors and the type of glycoconjugate on which they work. Several isozymes have been shown to work with small artificial acceptor substrates in in vitro assays or, in some cases, with large libraries of glycan acceptors. However, predictions of in vivo functions in cells based on these studies have often been erroneous, as found, for example, in the case of ST6GalNAc-II (Marcos, Bennett et al. 2011).

It is worth noting in this context that the deletion of glycosyltransferase genes using different siRNA silencing strategies has largely produced highly dubious information. In general, gene deletion in metabolic pathways where the encoded enzymes are often not limiting requires highly efficient deletion to produce discernible effects. Furthermore, siRNA technology is not 100% efficient, inevitably resulting in some heterogeneity of glycostructures. For the production of biopharmaceuticals, even a low percentage of an unwanted glycoform requires extensive analysis and monitoring.

Furthermore, glycosyltransferases are resident enzymes of the ER and Golgi with slow turnover, and it is very difficult to sufficiently reduce enzyme levels to affect glycosylation to a detectable degree, and it is essentially impossible to completely abrogate specific glycosylation capabilities with siRNA strategies in mammalian cells.

However, some studies have reported the effects of knocking down glycosylation capabilities with subsequent biological phenomena in cells despite residual enzyme levels (often confirmed by Western blots). Although some of these studies may have observed effects related to untargeted silencing strategies, it is clear that only complete deletion of the relevant glycosylation genes can unambiguously confirm these results, and examples of blatantly erroneous results have emerged. Deletion of the GALNT6 gene involved in protein O-glycosylation in a breast cancer cell line resulted in the loss of surface expression of the cancer-associated mucin MUC1 (Bennett, Mandel et al. 2012). However, MUC1 is expressed on the surface of many cells, including CHO cells, which do not express GALNT6, and therefore partial loss of the encoded GalNAc-T6 enzyme is unlikely to have such profound effects on MUC1 expression. Thus, it is clear that silencing screening strategies are not reliable for probing the in vivo cellular functions of glycosyltransferase genes to identify their role in glycosylation and to assess possible partial or complete functional redundancies in different glycosylation pathways.

Overexpression of glycosylation genes in cell lines

Furthermore, it should be noted in this context that transient or stable overexpression of a glycosyltransferase gene in a cell most often results in only partial changes in the glycosylation pathways in which the encoded enzyme is involved. Several studies have attempted to overexpress, for example, the core2 enzyme C2GnT 1 in CHO to produce core2 branched O-glycans, the sialyltransferase ST6Gal-I to produce α2,6-linked sialic acid on N-glycoproteins (E! Mai, Donadio-Andrei et al. 2013), and the sialyltransferase ST6GalNAc-1 to produce α2,6-linked sialic acid on O-glycoproteins forming the cancer-associated glycan STn (Sewell, Backstrom et al. 2006). However, all of these studies have resulted in heterogeneous and often unstable glycosylation patterns in transfected cell lines. This is presumed to be due in part to competitive activities of endogenous glycosyltransferases, whether acting on the same substrates or on substrates from divergent pathways. Other factors may also explain the heterogeneous glycosylation characteristics. This includes the availability of and competition for donor substrates and possible interference with other mechanisms required for the endogenous glycosylation machinery, including, for example, the subcellular topology and organization of glycosyltransferases, the coordinated functions of glycosyltransferases, as well as the availability of cofactors, chaperones, and other unknown factors required for efficient glycosylation.

Function of glycosylation in proteins

Protein glycosylation is an essential and highly conserved process in all mammalian cells, and the repertoire of glycosyltransferase genes available in mammalian cells is almost identical, with well-defined orthologous relationships elucidated (Hansen, Lind-Thomsen et al. 2014). Only a few glycosyltransferase genes have been inactivated through evolution in humans, such as the xenoantigen α3galactosyltransferase gene, GGTA1, and the Forssman α3GalNAc transferase gene, whereas humans have acquired some novel glycosyltransferase activities through the introduction of polymorphisms, such as the ABO blood group α3Gal/GalNAc transferase gene. Despite these minor differences in the glycosyltransferases that act on the end-capping of glycan structures, most glycosylation pathways in mammalian cells are identical and can be inferred from one mammalian cell to another regarding the functions of the expressed glycosyltransferases. Proteins entering the ER-Golgi secretory pathway in mammalian cells can undergo N-glycosylation as well as a variety of different types of O-glycosylation (Hansen, Lind-Thomsen et al. 2014). Proteins can also be cleaved and linked by a GPI anchor. Glycosylation affects proteins in several ways. Glycosylation can direct quality-controlled folding for ER exit and transport and secretion, and glycosylation can direct protein conformation and function, as well as sensitivity to proteolytic processing and/or degradation. Glycosylation is especially important for recombinant protein therapies produced in mammalian cells, and glycans can affect protein production, provide protein solubility and stability, affect protein bioactivity, biodistribution, and/or circulatory half-life, and target proteins to specific carbohydrate-binding receptors. Therefore, it is important to consider the glycosylation capacity of host cells used for recombinant therapeutic production for the production of effective therapies.

Recombinant production of therapies

Biological glycoproteins are the fastest growing therapeutic class (Walsh 2010), and most of them can only be produced recombinantly in mammalian cells with human-like glycosylation capacity. The Chinese hamster ovary cell has acquired a prominent role as a host cell for the recombinant production of therapeutic glycoproteins, mainly because it produces fairly simple N-glycans with branching and capping similar to those produced in some human cells (Walsh 2014). In particular, CHOs produce complex-type heterogeneous N-glycans with bi-, tri-, and tetra-antennary structures with an α6-fucose core, a minor amount of poly-N-acetyllactosamine (poly-LacNAc) mainly on the α1,6 arm of tetra-antennary structures, and a unique LacNAc cap with α2,3-neuraminic acid (NeuAc) (North, Huang et al. 2010). CHO does not generally produce immunogenic protective structures in non-humans and humans, such as N-glycolylneuraminic acid (NeuGc) or α3Gal, although their appearance can be a result of genetic induction (Bosques, Collins et al. 2010; North, Huang et al. 2010). N-glycosylation can vary for different proteins as well as for different glycosites within individual proteins, with IgG antibodies, for example, being produced with truncated N-glycan structures. A major concern with CHO is the substantial heterogeneity in N-glycan processing, which can be difficult to control during bioprocessing and pose bioactivity and biosafety concerns. Therefore, a major focus of therapeutic bioprocessing is on glycan analysis and fermentation control to achieve consistency (Walsh 2010).

Generic CHO cell lines produce N-glycans with a variable mixture of bi-, tri-, and tetra-antennary structures with a high degree of α6-fucosylation, and a lesser degree of poly-LacNAc structures and α2,3NeuAc capping on most glycoproteins (Sasaki, Bothner et al. 1987). Furthermore, some N-glycosylation sites on proteins may have incomplete stoichiometry. For IgG, CHO produces only bi-antennary N-glycans with low galactosylation and sialylation at the conserved site Asn297.

GalNAc O-glycans are produced with the core1 structure with α2,3NeuAc and partial α2,6-sialylation (αGalNAc) (Fukuda, Sasaki et al. 1989), and variable stoichiometries are also found. A CHO cell line with inactivated COSMC has been produced and this cell line produces only GalNAc O-glycan structures (Yang, Halim et al. 2014). Other types of O-glycans have not been studied, and only Fuc O-glycans are known to be elongated by LacNAc and capped with α2,3NeuAc.

There are no mammalian cell lines available that include CHO that can produce a single homogeneous antenna N-glycan species, neither biantennary N-glycans, triantennary N-glycans, nor tetraantennary N-glycans. Cells that do not produce small amounts of heterogeneous poly-LacNAc structures on antennal branches are also lacking. Furthermore, cells capable of producing homogeneous α2,6NeuAc-coated N-glycans are unavailable, and most human plasma proteins are coated with α2,6NeuAc rather than α2,3NeuAc as produced by CHO and many other mammalian cell lines.

Genetic engineering of CHO cells with respect to glycosyltransferase gene deletion is limited to single, isolated genes. There are no examples of stacked inactivation of multiple glycosyltransferase genes in mammalian cells, including CHO lines. Furthermore, there have been no examples of stacked inactivation of multiple glycosyltransferase genes encoding isozymes with potential related functions in protein glycosylation. Since many steps in protein glycosylation involve families of isozymes and their in vivo functions in cells are largely unknown, it is furthermore not possible to predict which genes to inactivate and/or introduce to obtain many of the preferred glycosylation capabilities for recombinant glycoprotein therapies.

Over the past two decades, significant efforts have been devoted to the genetic glycoengineering of CHO cells with the goals of expanding glycosylation capacity, reducing heterogeneity, and specifically enhancing or altering sialylation (Sinclair and Elliott 2005; Walsh and Jefferis 2006). These studies build on decades of work deciphering the biosynthetic pathways and genes involved in N-glycosylation (Kornfeld and Kornfeld 1985; Schachter 1991), and have essentially all used random integration of cDNAs encoding glycosyltransferases and have experienced problems with the stability, consistency, and predictability of the introduced glycosylation capacity (Kramer, Klausing et al. 2010). The major obstacle has been the need to rely on overexpression of glycosyltransferases with the resulting competition from endogenously expressed enzymes. Targeted gene knockdown in cell lines has been difficult to achieve in silencing strategies, but in the case of glycosyltransferase genes, success has been achieved with a strategy based on random mutagenesis in CHO followed by lectin-resistant selection (Patnaik and Stanley 2006). However, these mutant CHO cells generally do not have preferable glycosylation characteristics for recombinant therapies, may also contain other unknown mutations in the CHO genome, and none of them have been used for the production of clinical therapies.

The discovery of the importance of the central α6Fucosylation of the N-glycan Asn297 in IgG for antibody-dependent cellular cytotoxicity (ADCC) (Shinkawa, Nakamura et al. 2003) prompted the need to produce CHO cells with this glycosylation capacity, and fut8 knockout CHO lines are now available for IgG antibody production. In addition, CHO lines with knockout of either the ggta1 gene or the mgat1 gene have been generated.

However, CHO engineering has so far failed to address the most common problems of glycosylation heterogeneity, insufficient sialylation, and the introduction of novel glycosylation capabilities such as a more human-like sialylation (α2,6NeuAc). Currently, the bioprocessing industry for the recombinant production of glycoprotein therapeutics only has one generic CHO cell line with its natural glycosylation capability available for expression, with the exception of the aforementioned examples, of which only the fut8 knockout line is of value for classical therapeutics.

KANDA YUTAKA ET AL: "Comparison of cell lines for stable production of fucose-negative antibodies with enhanced ADCC", BIOTECHNOLOGY AND BIOENGINEERING, WILEY &SONS, HOBOKEN, NJ, US, vol.

94, no. 4, July 1, 2006 (2006-07-01), pages 680-688, XP002415543, ISSN: 0006-3592, DOI: 10.1002/BIT.20880 describes FUT8 knockout CHO cells, and their use for the production of fucose-negative therapeutic antibodies (abstract; page 681).

WONG ATHENA WET AL: "Enhancement of DNA uptake in FUT8- deleted CHO cells for transient production of afucosylated antibodies.", BIOTECHNOLOGY AND BIOENGINEERING 1AUG 2010, vol. 106, no. 5, August 1, 2010 (2010-08-01), pp. 751-763, XP002756439, ISSN: 1097-0290describes FUT8-deleted CHO cells for the production of afucosylated antibodies (abstract; pp. 753-755).

WO 2009/009086 A2 (SANGAMO BIOSCIENCES INC [USA]; COLLINGWOOD TREVOR [USA]; COST GREGORY J) January 15, 2009 (January 15, 2009) describes cell lines comprising an altered FUT8 gene, e.g., a deleted FUT8 gene, used for the production of recombinant proteins, such as, for example, antibodies. The cell line may be CHO, NSO, HEK293, or PER (abstract; paragraphs 7, 15-19, 31-33, 75, 122-125; example 4).

A. BUFFONE ET AL: "Silencing 1,3-Fucosyltransferases in Human Leukocytes Reveals a Role for FUT9 Enzyme during E-selectin-mediated Cell Adhesion", JOURNAL OF BIOLOGY AND CHEMISTRY, vol. 288, no. 3, November 28, 2012 (2012-11-28), pp. 1620-1633, XP55264669, USA, ISSN: 0021-9258, DOI:10.1074/jbc.M112.400929 describes FUT4, FUT7, and/or FUT9 knockout cells (abstract; pp. 1621, 21628-1629).

S. SHI TE AL: "Inactivation of the Mgat1 Gene ni Oocytes Impairs Oogenesis, but Embryos Lacking Complex and Hybrid N-Glycans Develop and Implant", MOLECULAR AND CELLULAR B iOLOGY., vol. 24, no. 22, November 15, 2004 (2004-11-15), pages 9920-9929, XP055264616, US, ISSN: 0270-7306, DOl: 10.1128/MCB.24.22.9920-9929.2004 describes deletions of the Mgat1 gene in oocytes for identify complex N-glycans (abstract; page 9922).

BHATTACHARYYA RIDDHI ET AL: "Biological consequences of inactivating the Mgat3 gene that encodes GIcNAc-TII", GLYCOBIOLOGY, vol. 10, no. 10, October 2000 (2000-10), page 1081, XP9189462, &5.8 ANNUAL CONFERENCE OF THE SOCIETY FOR GLYCOBIOLOGY; BOSTON, MASSACHUSETTS, USA; NOVEMBER 8-11, 2000, ISSN: 0959-6658describes cells comprising an inactivated Mgat3 gene.

WO 2013/106515 A1 (SIGMA ALDRICH CO LC [USA]; LNI NAN [USA]; SEALOVER NATALIE [USA]; GEORGE) July 18, 2013 (July 18, 2013) describes Mgat1-deficient cell lines. The cell line may be, for example, a CHO, BHK, HEK293 or NSO cell line. The gene could be inactivated. Inactivation of the Mgat1 gene is obtained by the action of a targeted endonuclease ZFN, TALEN,... Methods for preparing the Mgat1-deficient cell lines and for producing recombinant glycoproteins using such cell lines (abstract; paragraphs 4-6, 17, 18, 30-44, 79-81; examples 1, 3, 7).

Summary of the invention

The present invention relates to a mammalian cell in which mgat4A, mgat4B and mgat5 are deleted in the cell.

In another embodiment of the present invention, the cell is derived from Chinese hamster ovary or human kidney. In a further embodiment of the present invention, the cell is selected from the group consisting of CHO, NS0, SP2/0, YB2/0, CHO-K1, CHO-DXB11, CHO-DG44, CHO-S, HEK293, HUVEC, HKB, PER-C6, NS0 or derivatives of any of these cells. In one embodiment of the present invention, the cell is a CHO cell.

In another embodiment of the present invention, the cell further encodes an exogenous protein of interest. In yet another embodiment of the present invention, the protein of interest is an antibody, an antibody fragment, such as a Fab fragment, an Fc domain of an antibody, or a polypeptide.

In yet another embodiment of the present invention, the protein of interest is a coagulation factor such as coagulation factor II (FII), coagulation factor V (FV), coagulation factor VII (FVIIa), coagulation factor VIII (FVIII), coagulation factor IX (FIX), coagulation factor X (FX), or coagulation factor XIII (FXIII).

In a further embodiment of the present invention, the antibody is an IgG antibody.

In one embodiment of the present invention is the polypeptide erythropoietin (EPO), a protein involved in hemostasis, including a coagulation factor.

In one embodiment of the present invention, glycosylation is made more homogeneous.

In another embodiment of the present invention the glycosylation is non-sialylated.

In another embodiment of the present invention, the glycosylation is not galactosylated.

In one embodiment of the present invention it comprises the glycosylation of biantennary N-glycans.

In another embodiment of the present invention, the glycosylation does not comprise poly-LacNAc.

In yet another embodiment of the present invention glycosylation is any combination without fucose.

In a further embodiment of the present invention, FUT8 has been removed.

In one embodiment of the present invention, B4galt1 has been removed allowing for the generation of more homogeneous N-glycans without galactose.

In another embodiment of the present invention, B4galt1 and fut8 have been eliminated allowing the generation of more homogeneous N-glycans without galactose and fucose.

One aspect of the present invention relates to a method for producing a glycoprotein according to the present invention.

In one embodiment of the present invention, the cells have been modified by deleting the glycosyltransferase gene and/or inserting an exogenous DNA sequence encoding a glycosyltransferase. Another aspect of the present invention relates to a glycoprotein according to the present invention, which is a homogeneous glycoprotein conjugate comprising a polymer selected from PEG, HEP, XTEN, PSA, HES.

Brief description of the figures

Figure 1 shows the selection of ZFNs in CHO to define key glycosyltransferase genes involved in N-glycosylation using human EPO as a reporter. (a) Graphical representation of a common tetraantennary N-glycan with poly-LacNAc at the p6 antenna and sialic acid capping. Arrows indicate glycosyltransferase genes with potential roles in the biosynthesis of each step. Functional Glycomics Consortium designations of monosaccharides are indicated. (b) Glycolytic profile of EPO expressed in CHO GS WT cells. MALDI-TOF spectra of permetrified N-glycans released by PNGase F with labeling of four major species. (c) Glycoprofiling of EPO expressed in CHO GS cells with KO of genes involved in tri- and tetraantennary biosynthesis, showing that triple KO of mgat4A/4B/5 results in homogeneous biantennary N-glycans with less poly-LacNAc (lower panel). (d) Glycoprofiling with KO of p4galactosyltransferase genes involved in LacNAc biosynthesis, showing that only KO of B4galt1 alone resulted in a substantial but partial loss of galactosylation, whereas only stacked KO of B4galt1/3 resulted in almost complete loss of galactosylation. (e) Glycoprofiling with KO of two p3GlcNAc-transferase genes with claimed roles in poly-LacNAc biosynthesis, showing that KO of B3gnt2 results in complete loss of poly-LacNAc. (f) Glycoprofiling with KO of three α2,3-sialyltransferase genes with putative functions in N-glycosylation, showing that only the double KO of st3gal4/6 resulted in complete loss of sialic acid coverage.

Figure 2 shows the de novo engineering of sialylation capacity in CHO GS by ZFN-directed knockin. (a) Glycoprofiling of EPO expressed in WT, double st3gal4/6 KO, and st3gal4/6 KO with ST6GAL1 KI, showing that the range of tetraantennary N-glycan structures produced in WT CHO recover sialylation to the same extent after ST6GAL1 KI. (b) Glycoprofiling of EPO with triple mgat4A/4B/5 and B3gnt2 KO, showing complete loss of poly-LacNAc on biantennary N-glycans. (c) Glycoprofiling of mgat4A/4B/5 and st3gal4/6 KO without and with ST6GAL1 KI, showing complete loss of sialic acid capping on biantennary N-glycans and complete de novo capping by α2,6-sialic acid on biantennary N-glycans, respectively.

Figure 3 shows a graphical representation of the design matrix for genetically engineering the N-glycosylation capacity of CHO to produce homogeneous biantennary N-glycans with α2,3NeuAc cap, without sialylation, and with α2,3NeuAc cap on EPO. An additional knockout of fut8 would result in the loss of core fucosylation.

Figure 4 shows the expression profile of selected glycosyltransferase genes by RNAse in CHO. RNAse analysis was performed in two common CHO lines (CHO-K1, CHO-GS) and two independent mgat4A/4B/5 triple KO CHO GS clones (cl#1, cl#2). Included is the reported RNAse analysis of CHO K1, which was largely identical to our analyses except that the relative expression levels were higher. Importantly, some genes, including mgat4b, that were not expressed in CHO K1 were expressed in the present study, and mgat4b was found to be essential for the glycoengineering experiments described here. Another important observation was that the expression profiles of glycogens in the two triple KO clone analyses were identical to those of the parental CHO lines, providing evidence that precise gene editing does not alter the expression of other glycogens even for compensatory functional reasons. For a more conclusive assessment of this, we clearly need to obtain more extensive RNAseq data for all mutant clones.

Figure 5 shows the immunocytology of CHO knockout clones with mutations related to N-glycan branching and poly-LacNAc biosynthesis. (a) The lectin L-PHA was used to specifically probe branching (36) of the N-glycans and only the KO of mgat5 (mgat4A/4b/5) resulted in alterations in L-PHA labeling, whereas the control lectin ConA which binds all N-glycan types was unaffected. This confirms that MGAT4A and 4B form the p4 branch, whereas MGAT5 forms the p6 branch and thus supports the interpretation of the mass spectrometry data presented in Figure 1c with respect to branch assignments. (b) A similar strategy was used to investigate poly-LacNAc with the LEL, where only the KO of B3gnt2 resulted in the loss of labeling. These data also support the interpretation of the mass spectrometry data presented in Figure 2 with respect to the assignment of LacNAc to biantennary structures.

Figure 6 shows the immunocytology of CHO knockout clones with mutations linked to LacNAc formation. (a) MAb 1B2 was used to label for the presence of LacNAc after removal of sialic acids by neuraminidase pretreatment. Analysis of B4galt1-4 KO in heterogeneously branched WT CHO GS showed that B4galt1 KO reduced labeling, whereas only stacked B4galt1/3 KO completely abolished labeling. (b) The same analysis of B4galt1-4 KO in CHO GS engineered with homogeneous biantennary N-glycans (mgat4A/4B/5 KO) showed that KO of B4galt1 alone essentially abolished MAb 1B2 labeling, indicating that p4Gal-T1 is the primary enzyme responsible for galactosylation of biantennary N-glycans. This finding is expected to have implications for the recombinant production of therapeutic IgGs in CHO, as these proteins are generally produced with galactosylation-limited biantennary N-glycans.

Figure 7 shows the immunocytology of sialylation-related CHO knockout and knockin clones. (a) MAb 1B2 was used to label LacNAc exposure resulting from the loss of sialylation ability, and this showed that KO of st3gal3 and st3gal6 did not substantially affect labeling, whereas KO of st3galt4 and stacked KO of st3galt4/6 resulted in strong labeling. This suggests that the ST3Gal-IV enzyme is the major sialyltransferase responsible for the α2,3-sialylation of N-glycans. Furthermore, de novo introduction of human ST6Gal-1 into the stacked st3galt4/6 KO clone completely abolished MAb 1B2 labeling, demonstrating efficient sialylation by this enzyme. (b) To confirm that de novo introduction of human ST6Gal-1 resulted in α2,6-sialylation, we used the lectin SNA, which did not label WT CHO cells as expected and only labeled after ST6GAL1 introduction. This result supports the mass spectrometry interpretations presented in Figure 2.

Figure 8 shows the analysis of targeted knockin of ST6GAL1 by ligation PCR. (a) We used a modified Obligare targeted KI strategy using two inverted ZFN binding sites flanking the ST6GAL1 gene in the donor plasmid. 5' and 3' ligation PCR confirmed targeted integration at the Safeharbor#1 site in the CHO clone with st3gal4/6 KO. (b) 5' and 3' ligation PCR confirmed targeted integration in the CHO clone with st3gal4/6/mgat4A/4B/5 KO. The allelic copy number status of integration was determined by WT allelic PCR. The presence of the desired band in the WT allelic PCR indicates the presence of the Safeharbor#1 site without integration of the targeted KI of the gene of interest in at least one of the alleles. The results showed biallelic integration of st6gal1 to the Safeharbor#1 site in st3gal4/6 KO and monoallelic integration to mgat4A/4B/5/st3gal4/6 KO, respectively.

Figure 9 shows the Coomassie SDS-PAGE analysis of purified recombinant EPO expressed in CHO clones. Approximately 1 μg of purified EPO was loaded based on OD.

Figure 10 shows glycoprofiling of EPO produced in CHO-GS WT, B3gnt2/mgat4A/4B/5, and st3gal4/6/mgat4A/4B/5 KO with ST6Gal-I KI, showing complete loss of poly-LacNAc on the biantennary N-glycans, and complete de novo capping by α2,6-sialic acid on the biantennary N-glycans without poly-LacNAc. Negative ion ESI-MS glycoprofiling was used to validate the MALDI-TOF-based glycan profile data.

Figure 11 shows negative ion ESI-MS/MS of the precursor ions at m/z 1183.92 (sialylated biantennary N-glycan with central fucose) of EPO produced in CHO with (A) mgat4A/4B/5 KO and with (B) both mgat4A/4B/5 and st3gal4/6 KO and KI of ST6Gal-I. The presence of the diagnostic fragment ions at m/z 306.12 indicates a terminal α2,6 sialylation.

Figure 12. A graphical representation of the human CAZ and GT families involved in the biosynthesis of major mammalian glycoconjugates and their different glycosylation pathways. Common mammalian glycan structures are depicted for (A) O-glycans including O-GalNAc, O-GlcNAc, O-Fuc, O-Glc, O-Man, O-Xyl; (B) N-glycans; (C) GPI anchors; (D) C-glycans; (E) glycospingolipids; (F) hyaloronane and (G) hydroxy-lysine. Functional Glycomics Consortium designations of monosaccharides are indicated.

Figure 13 A graphical representation of the different glycan structures and glycosylation pathways found in CHO cell lines with designations of glycosyltransferase genes expressed in CHO cell lines and their predicted role in biosynthetic steps. Genes are assigned to major confirmed or putative functions in O-glycosylation (O-GalNAc (A1), O-Man (A2) and O-Fuc, O-Glc, O-GlcNAc, O-Xyl (A3)) and N-glycosylation pathways relevant to recombinant glycoprotein therapeutics today. (A1-A4) O-GalNAc glycosylation pathway with extended structures of core 1, core 2, core 3 and core 4; (B) N-glycosylation pathway; (C) Glycosphingolipid biosynthesis pathways; (D) C-glycan, and (E) GPI-anchors. CAZy families of glycosyltransferases (GT genes) are annotated, and GT genes expressed in CHO cell lines are shown. Glycan structures not synthesized in CHO are boxed. Functional Glycomics Consortium designations for monosaccharides are indicated.

Figure 14 Graphical representations of all genes encoding isozymes with potential to regulate N-glycosylation branching, elongation, and end-capping expressed in CHO. The representation shows the knockout screen of glycosyltransferase genes involved in N-glycosylation in CHO using human EPO as a recombinantly expressed reporter glycoprotein. (A) The common tetraantennary N-glycan with poly-LacNAc on the p6 antenna and capped by sialic acids is shown, knockout genes are boxed, and genes not expressed in CHO cells are annotated. Note that human ST6Gal-I (ST6GAL1) was introduced by ZFN-mediated knock-in. Functional Glycomics Consortium (FGC) designations of the monosaccharides are indicated. (B) Schematic representation of the actual target regions of the nucleases relative to the predicted overall domain structure of type II glycosyltransferase proteins (top panel) and the target exon for each gene by numbering the respective genes starting from the first coding exon (bottom panel). The exon structure representation only includes the first selected exons and does not include all exons from all genes.

Figure 15 Glycoprofile of EPO expressed in WT CHO cells. PNGase F MALDI-TOF spectra revealed permethylated N-glycans with predicted structures for the four major species.

Figure 16 Glycoprofiling of EPO expressed in CHO cells with KO of genes involved in complex N-glycan biosynthesis and antennae formation, showing that the double KO of mgat4B/5 (B) and the triple KO of mgat4A/4B/5 (B) result in homogeneous biantennary N-glycans with a lower amount of poly-LacNAc. Since the single KO of mgat4A (A) had minor effects compared to WT, it is likely that lower amounts of triantennary structures are present in the double KO of mgat4B/5.

Figure 17 Glycoprofiling with KO of the B4galt1/2/3/4 genes involved in LacNAc biosynthesis, showing that only the B4galt1 KO alone resulted in a substantial but partial loss of galactosylation, whereas only the stacked B4galt1/3 KO resulted in an almost complete loss of galactosylation. The stacked B4galt1/2/3 KO resulted in an essentially complete loss of galactosylation.

Figure 18. Glycosylation profile of three p3GlcNAc transferase genes showing that B3gnt2 knockout results in complete loss of poly-LacNAc, whereas knockout of B3gnt1 and B3gnt1 had no effect. The bottom panel shows that knockout of B3gnt2 in combination with mgat4A/4B/5 results in complete loss of poly-LacNAc in biantennary structures.

Figure 19 Glycoprofiling with KO of three α2,3-sialyltransferase genes with claimed roles in N-glycosylation, showing that only the double KO of st3gal4/6 resulted in complete loss of the sialic acid coating.

Figure 20 Immunofluorescence cytology with ConA and L-PHA lectin staining showing loss of L-PHA with mgat5 deletion.

Figure 21 Immunofluorescence cytology with a monoclonal antibody (clone 1B2) against LacNAc. Panel (A) shows, surprisingly, that only stacked KO of B4galt1/3, but not B4galt1/2/4, abolished galactosylation. Panel (B) shows that KO of B4galt1 in CHO-GS with KO of mgat4A/4B/5 abolished LacNAc immunoreactivity, whereas KO of B4galt2, 3, and 4 in the same CHO-GS had no substantial effects.

Figure 22 Recombinant expression of a therapeutic IgG in B4galt1 knockout CHOs resulted in homogeneous biantennary N-glycans without galactosylation. Glycoprofiling of IgG produced in WT and B4galt1 KO CHOs shows essentially complete loss of the characteristic incomplete galactosylation and sialylation of the conserved N-glycan at Asn297. B4galt1 knockout in combination with fut8 resulted in N-glycans without galactosylation and fucose.

Figure 23 LEL lectin immunofluorescence cytology shows that B3gnt2 is the key gene controlling LacNAc initiation in CHO cells.

Figure 24 Immunofluorescence cytology with monoclonal antibody against LacNAc (1B2). Knockdown of St3gal3, St3gal4, and St3gal6 individually by immunocytology to assess LacNAc exposure shows that only St3gal4 knockout resulted in substantial LacNAc exposure.

Figure 25 Glycoprofiling of EPO expressed in mgat4a/4b/5 knockout CHOs and ZFN-mediated knockin of human ST6GAL1. Panel A shows the EPO profile with homogeneous α2,6NeuAc coating of the full range of antennal N-glycan structures. Also shown is ST6GAL1 KI in mgat4a/4b/5/B3gnt2 knockout CHOs, which produced more homogeneous tetra-antennary structures. Panel B shows the profile with additional mgat4a/4b/5 knockout where EPO is produced with homogeneous biantennary N-glycans covered by α2,6NeuAc. Human ST3Gal-4 KI in combination with st3gal4/6/mgat4a/4b/5/B3gnt2 KO are also shown where EPO is produced with homogeneous biantennary N-glycans without poly-LacNAc and capped by α2,3NeuAc.

Figure 26 Negative ion ESI-MS/MS of precursor ions at m/z 1183.92 (sialylated biantennary N-glycan with a fucose core) from EPO produced in CHO with (A) Mgat4a/4b/5 KO and with (B) both Mgat4a/4b/5 and st3gal4/6 KO and KI of ST6GAL1. The presence of diagnostic fragment ions at m/z 306.12 demonstrates the presence of α2,6-terminal sialylation.

Figure 27 Targeted KI analysis of ST6GAL1 by ligation PCR. Panel A shows that a modified ObLiGaRe-targeted KI strategy was used using two inverted ZFN binding sites flanking the complete ST6GAL1 open reading frame in the donor plasmid. 5' and 3' ligation PCR confirmed targeted integration at the Safeharbor#1 site in the CHO clone with st3gal4/6 KO. (b) 5' and 3' ligation PCR confirmed targeted integration in the CHO clone with st3gal4/6/mgat4a/4b/5 KO. The allelic copy number status of integration was determined by WT allelic PCR. The presence of the desired band in the WT allelic PCR indicates the presence of the Safeharbor#1 site without integration of the targeted KI of the gene of interest in at least one of the alleles. The results showed biallelic integration of ST6Gal-I at the Safeharbor#1 site in the CHO st3gal4/6 KO clone and monoallelic integration in the mgat4a/4b/5/St3gal4/6 KO clone. The ObLiGaRe KI strategy was highly efficient, with approximately 30–50% of individual cloned cells expressing ST6GAL1, as assessed by immunocytology with antibodies and lectins.

Figure 28 Graphical representation of the genetic deconstruction of N-glycosylation in CHO established using EPO as an N-glycoprotein reporter. Several KO CHO lines have potential for the production of therapeutic glycoproteins with more homogeneous glycosylation (e.g., α2,3-sialylated biantennary N-glycans for EPO, and non-galactosylated/sialylated biantennary with and without Fuc core for IgG). Examples of KI-based reconstructions are shown producing homogeneous α2,6-sialylation after KO of α2,3-sialylation capabilities with WT N-glycan branching heterogeneity or homogeneous biantennary structure, homogeneous α2,3-sialylation after KO of endogenous sialylation, and homogeneous tetraantennary structures with α2,6-sialylation after KO of endogenous branching and sialylation capabilities. The reconstruction can address homogeneity as shown, but also branching, poly-LacNAc elongation, and any type of protection as indicated.

Figure 29 Glycoprofiling of EPO expressed in CHO with Cosmc knockout to eliminate O-GalNAc elongation with Gal and NeuAc showing an enhanced ability for sialic acid coating of N-glycans at all three N-glycosites present in EPO, compared to WT as shown in Figure 15.

Figure 30 Glycoprofiling of CHO-expressed EPO knocked out of B4galT7 to eliminate O-Xyl elongation with Gal (PANEL A) and Pomgnt1 to eliminate O-Man elongation with Gal and NeuAc (PANEL B) showing enhanced ability for sialic acid capping of N-glycans at all three N-glycosites present on EPO. PANEL C illustrates glycoprofiling of EPO expressed in double KOs of B4galt7 and pomgnt1, and Panel D illustrates glycoprofiling of EPO expressed in triple KOs of B4galt7, pomgnt1, and COSMC showing the same enhanced ability for sialic acid capping of N-glycans on EPO.

Figure 31 Glycoprofiling of EPO expressed in CHO with B4galT4 knockout to remove an unused galactosyltransferase paralog for N-glycans and this glycosylation pathway shows an enhanced ability to coat N-glycans with sialic acid at all three N-glycosites present in EPO.

Figure 32 Glycoprofiling of EPO expressed in mgat2 knockout CHO showing loss of tetra- and triantennary N-glycan structures and the appearance of mono- and biantennary N-glycans with galactose and NeuAc protection (PANEL B), and without sialic acid protection in st3gal4/6 knockout (PANEL C). Further deletion of mgat4A/4B/5 resulted in monoantennary structures with lower amounts of poly-LacNAc (PANELS D and E). Reintroduction of human mgat4A and mgat5 in combination with ST6Gal-I resulted in the loss of monoantennary N-glycans and the restoration of tri- and tetraantennary structures (PANELS H and I).

Figure 33 SDS-PAGE gel analysis of glycoPEGylation reactions of recombinant wild-type EPO-Myc-His6 and recombinant EPO-Myc-His6 with single-antenna N-glycans. Lanes contained: (1) HiMark Standard, (2) ST3Gal3, (3) Sialidase, (4) wt hEPO, wt EPO-Myc-His6 glycoPEGylated with ST3Gal-III and (5) 1x, (6) 5x, (7) 10x, (8) 25x, (9) 50x 10kDa-PSC reagent respectively, (10) EPO-Myc-His6 glycosylation variant and EPO-Myc-His6 glycoPEGylated variant with ST3Gal-III and (11) 1x, (12) 5x, (13) 10x, (14) 25x, (15) 50x 10kDa-PSC reagent. Bands above approximately 160 kD indicate glycopegylation with more than two PEG chains on one or more N-glycans.

Figure 34 Quantification of the number of linked PEG chains by densitometry of SDS-PAGE gels. (PANEL A) Product profile of the enzymatic reaction of 10 kDa glycoPEGylation with EPO-Myc-His6 produced in ordinary CHO cell lines. (PANEL B) Product profile of the enzymatic reaction of 10 kDa glycoPEGylation with EPO-Myc-His6 produced in Mgat2/4a/4b/5 KO CHO cell lines. Profiles A and B were obtained by densitometric measurements in lane 9 and lane 15 of Figure 33, respectively.

Figure 35 Glycoprofile of EPO expressed in B3gnt2 knockout CHO cells stacked with Cosmc/Pomgnt1/B4galt7 knockout (the latter triple knockout seen in Fig. 30D) (top panel), B4galt5 and B4galt6 double knockout (second panel), Mgat2 and Mgat4B double knockout (third panel), and Mgat2 and Mgat5 double knockout (bottom panel). Compared to background (see Fig. 30D), B3gnt2 knockout deletes polyLacNac. B4galt5 and B4galt6 knockout yields more LacNac compared to wt EPO (Fig. 31). The last two panels with double KO of Mgat2/4b or Mgat2/5 show that monoantennary can be obtained with double KO, so KO of four Mgat2/4A/4B/5 genes is not necessary (compare with Fig. 32D).

Figure 36 Glycoprofile of IgG expressed in CHO cells with Mgat2 KO (top), St3Gal4/6 double KO (middle), or B4galt1/3 KO stacked on Cosmc/Fut8 (bottom panel). Mgat2 KO results in a homogeneous mono-antennal structure and a higher degree of sialylation than wt IgG (compare with Fig. 22). St3gal4/6 KO gives higher galactosylation (compare with Fig. 22). B4galt1 and B4galt3 double KO show more homogeneous galactose-free N-glycans than those obtained with B4galt1 single KO (compare with Fig. 22, bottom panel).

The present invention will be described in more detail below.

Detailed description of the invention

The sugar chains of glycoproteins are broadly divided into two types, namely, a sugar chain that is linked to asparagine (sugar chain linked to an N-glycoside) and a sugar chain that is linked to another amino acid such as serine, threonine (sugar chain linked to an O-glycoside), depending on the mode of attachment to the protein moiety.

The end of the sugar chain that is linked to asparagine is called the reducing end, and the opposite side is called the non-reducing end. It is known that the sugar chain linked to N-glycosides includes a high-mannose type in which mannose alone is linked to the non-reducing end of the core structure; a complex type in which the non-reducing end of the core structure has at least one parallel galactose-N-acetylglucosamine branch (hereinafter referred to as "Gal-GlcNAc") and the non-reducing end of Gal-GlcNAc has a sialic acid structure, N-acetylglucosamine bisection, or the like; and a hybrid type in which the non-reducing end of the core structure has branches of both the high-mannose type and the complex type.

In general, most humanized antibodies being considered for drug application are prepared using genetic recombination techniques and produced using CHO cells derived from Chinese hamster ovary tissue as the host cell. As described above, the sugar chain structure plays an important role in the structure, function, size, circulatory half-life, and pharmacokinetic behavior of glycoprotein drugs. Furthermore, the sugar structure plays a notably important role in the effector function of antibodies, and differences in the sugar chain structure of glycoproteins expressed by host cells are observed. Therefore, the development of a host cell that can be used for the production of an antibody with enhanced effector function is desired.

To support the production of an increasing number of biopharmaceutical glycoproteins, it is necessary to develop new mammalian cell lines and, preferably, CHO-derived cell lines with different glycosylation capacities and characteristics.

Furthermore, there is a need to develop design matrices for individual glycosylation pathways that involve gene inactivation and/or the introduction of stable genes that allow the generation of tailored mammalian cell lines with different and more homogeneous glycosylation capabilities and characteristics.

Furthermore, there is a need to develop mammalian cell lines, and preferably CHO-derived cell lines, with inactivation of two or more glycosyltransferase genes that can produce recombinant glycoproteins with glycosylation different from that of their natural counterpart. More particularly, there is a need to develop such cell lines with inactivation of two or more glycosyltransferase genes that encode isoenzymes with related functions in the same glycosylation pathway, for example the N-glycosylation pathway. Furthermore, there is a need to develop such cell lines with inactivation of two or more glycosyltransferase genes that encode enzymes with unrelated functions in the same glycosylation pathway.

Furthermore, there is a need to introduce one or more glycosyltransferases into cell lines, preferably CHO-derived cell lines, to obtain desirable glycosylation capabilities, including homogeneous and novel capabilities. This introduction of glycosyltransferases can be combined with the inactivation of one or more of the endogenous glycosyltransferase genes.

An object of the present invention is to provide a modified cell with stable capacity genetically designed for the production of a therapeutic protein, and which produces said protein with glycan chains with defined structures, and/or more homogeneous glycan structures, and/or with improved bioactivity.

This invention discloses a glycosyltransferase gene knockout and deconstruction system in CHO that provides a design matrix for engineering a cell with a multitude of well-defined glycosylation capabilities. Furthermore, the invention provides reconstructive designs for de novo engineering of a multitude of well-defined desired and/or novel glycosylation capabilities by combining one or more glycosyltransferase gene knockout and introduction events into a cell to enhance the production of recombinant glycoprotein therapeutics.

This invention discloses a mammal in which mgat4A, mgat4B and mgat5 are deleted in the cell, and with novel glycosylation capabilities that allow improvements for recombinant production of therapeutic proteins.

In one aspect, ZFN targeting designs for glycosyltransferase gene inactivation are provided.

In another aspect, TALEN designs for the inactivation of glycosyltransferase genes are provided. In yet another aspect, CRISPR/Cas9-based targeting for the inactivation of glycosyltransferase genes is provided.

The present inventors have employed for the first time a ZFN-mediated knockout screen in CHO to explore the potential for engineering N-glycosylation of recombinant glycoproteins. Many steps in the N-glycan biosynthetic pathway are potentially catalyzed by multiple isozymes, making genetic engineering unpredictable (Fig. 1a). Analyzing the in vivo functions of each of these isozymes is necessary to construct a matrix of design options.

The present inventors have examined the genes involved in this highly complex machinery (see examples) and have identified effects.

Genes can be deleted (KO) or included (KI).

In one aspect of the present invention, mgat4A/4B is deleted in the cell to remove branched tetraantennary N-glycans on p4 and mgat5 is deleted in the cell to remove branched tetraantennary structures on p6.

In yet another aspect of the present invention, mgat5 is also deleted in the cell, leading to the loss of L-PHA lectin labeling.

In a further aspect of the present invention, mgat4A, mgat4B and mgat5 are deleted in the cell, leading to homogeneous biantennary N-glycans.

In a further aspect of the present invention, B3gnt2 is also deleted in the cell, leading to the removal of poly-LacNAc.

In another aspect of the present invention, st3gal4, st3gal6, mgat4A, mgat4B and mgat5 are deleted in the cell, resulting in biantennary N-glycans without sialylation, but increased poly-LacNAc.

In another aspect of the present invention, st3gal3, st3gal4 and st3gal6 are also deleted in the cell, leading to a complete lack of sialic acid.

In a further aspect of the present invention, st3gal4, st3gal6, mgat4A, mgat4B, mgat5 are deleted in the cell, leading to homogeneous biantennary N-glycans covered by α2,6NeuAc.

Introduction of ST6GAL1 (knock in, KI) with KO of st3gal4/6 produces the full range of N-glycan antennal structures with a normal degree of NeuAc coating (Fig. 2a).

ST6GAL1 introduced with additional KO of mgat4A/4B/5 produces glycoproteins with homogeneous biantennary N-glycans capped by α2,6NeuAc (Fig. 2b).

Reintroduction of ST6GAL1 abolishes the small amounts of poly-LacNAc formed in biantennary structures when sialylation is removed (Fig. 2b).

Therefore, the ST6GAL1 KI can be used in combination with one or more of the knockouts described herein.

In the present context, NeuAc is also known as neuraminic acid.

Therefore, one aspect of the present invention relates to a protein expressed in a cell, wherein the post-translational modification pattern comprises homogeneous biantennary N-glycans with or without α2,3NeuAc protection or with and without α2,6NeuAc protection.

Deletion refers to the total or partial alteration of the function of the gene of interest.

In one aspect of the present invention, one or more of the aforementioned genes are deleted using ZFN zinc finger nucleases. ZFNs can be used to inactivate a FUT8 gene or any of the other genes described herein. ZFNs comprise a zinc finger protein (ZFP) and a nuclease (cleavage) domain.

In another embodiment, the present invention provides CHO cell lines with different well-defined N-glycosylation capabilities that allow for the recombinant production of therapeutic glycoproteins with N-glycans composed of biantennary N-glycans with or without poly-LacNAc, and with or without α2,3NeuAc coating.

In another embodiment, this invention provides cell lines with different well-defined N-glycosylation capabilities that allow for the recombinant production of therapeutic glycoproteins with N-glycans composed of biantennary, triantennary, or tetraantennary N-glycans with or without poly-LacNAc, and with or without α2,6NeuAc coating.

In yet another embodiment, this invention provides cell lines that allow for the production of glycoproteins with homogeneous N-glycans without sialic acid protection, allowing for direct enzymatic modification of N-glycans by sialyltransferases in vitro after production.

In yet another embodiment, the present invention provides cell lines that allow for the production of glycoproteins with homogeneous biantennary state N-glycans without sialic acid coating, allowing for direct enzymatic modification of the N-glycans by sialyltransferases post-production in vitro.

In another embodiment, the present invention provides cell lines that allow the production of glycoproteins with homogeneous mono-antennary state N-glycans with or without sialic acid coating, allowing direct enzymatic modification of an N-glycan by sialyltransferases in vitro post-production.

In yet another aspect, there is also provided an isolated cell comprising any of the proteins and/or polynucleotides as described herein. In certain embodiments, one or more glycosyltransferase genes are inactivated (partially or completely) in the cell. Any of the cells described herein may include additional genes that have been inactivated, for example, using zinc finger nucleases, TALENs, and/or CRISPR/Cas9 designed to bind to a target site on the selected gene. In certain embodiments, provided herein are cells or cell lines in which two or more glycosyltransferase genes have been inactivated, and cells or cell lines in which one or more glycosyltransferase genes have been inactivated and one or more glycosyltransferases have been introduced.

In yet another embodiment, this invention provides cell lines with inactivation of undesirable glycosylation pathways for optimized production of glycoproteins with desirable glycosylation characteristics. In certain embodiments, inactivation of one or more biosynthetic steps in protein O-glycosylation pathways (O-GalNAc, O-Xyl, O-Man, and glycolipid) is achieved individually or in combination to enhance the pool of sugar nucleotides available for desirable glycosylation characteristics.

In one embodiment, this invention provides a cell with mgat2, mgat4A, mgat4B, and mgat5 inactivation, producing erythropoietin with N-glycans that have homogeneous monoantennary structures, and are suitable for more homogeneous glycomodification.

For example, the cell lines described herein that have inactivated the mgat2, mgat4A, mgat4B, mgat5, st3gal4 and 6 genes, produce erythropoietin with N-glycans that have homogeneous monoantennary structures without sialic acid coating, and are suitable for more homogeneous glycomodification.

Inactivation of unnecessary glycosyltransferases to improve desirable glycosylation pathways. In yet another embodiment, this invention provides cell lines with inactivation of undesirable glycosylation enzymes to improve the capacity and/or fidelity of desirable glycosylation characteristics. In certain embodiments, cell lines are provided with activation of one or more sialyltransferases, and/or galactosyltransferases, and/or GlcNAc-transferases, that do not function in a desirable glycosylation pathway, such as for example N-glycosylation.

In one embodiment, this invention provides a cell with inactivation of a p4-galactosyltransferase (p4Gal-T7) with little or no function in the N-glycosylation pathway, and improved capacity for galactosylation, polyLacNAc, and sialic acid capping of N-glycans.

In one embodiment, this invention provides a cell with inactivation of a sialyltransferase with little or no function in the N-glycosylation pathway, and improved capacity for galactosylation and sialic acid coating of N-glycans.

In one embodiment, this invention provides a cell with inactivation of a p3galactosyltransferase (C1GalT1) with little or no function in the N-glycosylation pathway, and improved capacity for galactosylation and sialic acid capping of N-glycans.

In one embodiment, this invention provides a cell with inactivation of a p4galactosyltransferase with little or no function in the N-glycosylation pathway, and improved capacity for galactosylation and sialic acid coating of N-glycans.

In one embodiment, this invention provides a cell with inactivation of a p2GlcNAc-transferase (POMGnT1) with little or no function in the N-glycosylation pathway, and improved capacity for galactosylation and sialic acid capping of N-glycans.

In one embodiment, this invention provides a cell with inactivation of a p3galactosyltransferase (C1GalT1), and/or a p2GlcNAc-transferase (POMGnT1), and/or a p4galactosyltransferase (p4Gal-T7) with little or no function in the N-glycosylation pathway, and improved capacity for galactosylation and sialic acid coating of N-glycans.

In one embodiment, this invention provides a cell with inactivation of a p4galactosyltransferase (p4Gal-T4) with little or no function in the N-glycosylation pathway, and improved capacity for galactosylation and sialic acid coating of N-glycans.

Additionally, methods of using the zinc finger proteins and fusions thereof in methods of inactivating glycosyltransferase genes in a cell or cell line are provided. In certain embodiments, inactivating one or more glycosyltransferase genes results in a cell line that can produce recombinant proteins at higher levels or in which one or more activities (functions) of the proteins are increased compared to proteins produced in cells in which the gene(s) are not inactivated.

Thus, in another aspect, provided herein are methods for inactivating at least the cellular glycosyltransferase genes mgat4A, mgat4B, and mgat5 (e.g., also the endogenous genes mgat1, mgat2, mgat3, mgat4A, mgat4B, mgat4C, mgat5, mgat5B, B4galt1, B4galt2, B4galt3, B4galt4, B3gnt1, B3gnt2, B3gnt8, st3gal3, st3gal4, st3gal6, fut8) in a cell, by using methods comprising genome perturbation, gene editing, and/or gene disruption capability, such as clustered regularly interspaced short palindromic repeats (CRISPR)-related nucleic acid vector systems and their components, nucleic acid vector systems encoding fusion proteins comprising binding domains to zinc finger (ZF) DNA and at least one cleavage domain or at least one cleavage half-domain (ZFN) and/or nucleic acid vector systems encoding a first transcription activator-like effector endonuclease (TAL) monomer and a nucleic acid encoding a second cleavage domain or at least one cleavage half-domain (TALEN). The introduction into a cell of any of the aforementioned nucleic acid cleavage agents (CRISPR, TALEN, ZFN) are capable of specifically cleaving a glycosyltransferase gene target site as a result of the cellular introduction of: (1) a nucleic acid encoding a pair of glycosyltransferase gene target binding proteins ZF or TAL, each fused to said nucleic acid cleaving moiety, wherein at least one of said ZF or TAL polypeptides is capable of specifically binding to a nucleotide sequence located upstream of said target cleavage site, and the other ZF or TAL protein is capable of specifically binding to a nucleotide sequence located downstream of the target cleavage site, whereby each of the zinc finger proteins independently binds to and surrounds the nucleic acid target, followed by disruption of the target nucleic acid by double-strand breaks mediated by the fused endonuclease cleaving molecules, (2) a nucleotide sequence encoding a guide RNA of the CRISPR-Cas system that hybridizes to the target sequence of the glycosyltransferase gene, and b) a second nucleotide sequence encoding a type II Cas9 protein, wherein components (a) and (b) are located in the same vector or in different vectors of the system, wherein the guide RNA is composed of a chimeric RNA and includes a guide sequence and a cr trans-activator sequence (tracr), whereby the guide RNA is directed to the target sequence of the glycosyltransferase gene and the Cas9 protein cleaves the target site of the glycosyltransferase gene.

In yet another embodiment, this invention provides cell lines with inactivation of one or more glycosyltransferase genes and with stable introduction of one or more glycosyltransferases to improve the fidelity of desirable glycosylation characteristics and/or introduce improved glycosylation characteristics and/or novel glycosylation characteristics.

In certain embodiments, cell lines are provided with inactivation of one or more sialyltransferases, and/or galactosyltransferases, and/or glucosyltransferases, and/or GlcNAc-transferases, and/or GalNAc-transferases, and/or xylotransferases, and/or glucuronosyltransferases, mannosyltransferases, and/or fucosyltransferases, and into which one or more glucosyltransferases have been stably introduced.

For example, cell lines such as those described herein that have inactivated mgat4A, mgat4B, mgat5, st3gal4 and 6 genes and into which the human sialyltransferase ST6Gal-I has been introduced, produce erythropoietin with biantennary N-glycans homogeneously covered by α2,6NeuAc and are more similar to humans.

For example, the cell lines described herein that have inactivated the mgat4A, mgat4B, mgat5, st3gal4 and 6 genes and into which the human MGAT4A and ST6Gal-I genes have been introduced, produce erythropoietin with triantennary N-glycans homogeneously capped by α2,6NeuAc and are more similar to humans.

For example, the cell lines described herein that have inactivated the mgat4A, mgat4B, mgat5, st3gal4 and 6 genes and into which MGAT4A, MGAT5 and ST6Gal-I have been introduced, produce erythropoietin with tetraantennary N-glycans homogeneously capped by α2,6NeuAc and are more similar to humans.

There are several methods available for introducing exogenous genes, such as glycosyltransferase genes, into mammalian cells and selecting stable clonal cell lines that harbor and express the gene of interest. Typically, the gene of interest is co-transfected with a selectable marker gene that favors cell lines expressing the selectable marker under certain defined media culture conditions. The medium may contain an inhibitor of the selectable marker protein, or the medium composition may stress cellular metabolism, thus requiring increased expression of the selectable marker. The selectable marker gene may be present on the same plasmid as the gene of interest or on another plasmid.

In one embodiment, the introduction of one or more exogenous glycosyltransferases is carried out by plasmid transfection with a plasmid encoding constitutive promoter-driven expression of both the glycosyltransferase gene and an antibiotic selectable marker, where the selectable marker could also represent an essential gene not present in the host cell, such as the GS system (Sigma/Lonza), and/or separate plasmids encoding either the constitutive promoter-driven glycosyltransferase gene or the selectable marker. For example, plasmids encoding ST6GalNAc-I and Zeocin have been transfected into cells, and stable lines expressing ST6Gal-I have been selected based on Zeocin resistance.

In another embodiment, the introduction of one or more exogenous glycosyltransferases is accomplished by site-directed nuclease-mediated insertion.

In one embodiment, a method is provided for stably expressing at least one product of an exogenous nucleic acid sequence in a cell by introducing double-strand breaks into the PPP1R12C or Safe Harbor#1 genomic locus using ZFN nucleic acid cleavage agents and an exogenous nucleic acid sequence that, through a homology-dependent manner or via compatible flanking ZFN cleavage overhangs, is inserted into the cleavage site and expressed. Safe Harbor sites are sites in the genome that, when manipulated, have no apparent cellular or phenotypic consequences. In addition to the sites mentioned above, several other sites have been identified, such as Safe Harbor#2, CCR5, and Rosa26. In addition to ZFN technology, TALEN and CRISPR tools can also allow for the integration of exogenous sequences into cell lines or genomes in a precise manner. This requires assessing: i) the extent to which epigenetic silencing occurs, and ii) the desired expression level of the gene of interest. In the examples included herein, site-specific integration of human glycosyltransferases, e.g., ST6Gal-I, MGAT4A, or MGAT5, using the ObLiGaRe insertion strategy, is based on an insulator-flanked vector design driven by CMV expression.

In yet another aspect, the disclosure provides a method for producing a recombinant protein of interest in a host cell, the method comprising the steps of: (a) providing a host cell comprising two or more endogenous glycosyltransferase genes; (b) inactivating the endogenous glycosyltransferase genes of the host cell by any of the methods described herein; and (c) introducing an expression vector comprising a transgene, the transgene comprising a sequence encoding a protein of interest, into the host cell, thereby producing the recombinant protein. In certain embodiments, the protein of interest comprises, for example, erythropoietin or an antibody, for example, a monoclonal antibody.

In yet another aspect, the disclosure provides a method for producing a recombinant protein of interest in a cell, the method comprising the steps of: (a) providing a cell according to the present invention; (b) inactivating the endogenous glycosyltransferase gene(s) of the host cell; (c) introducing one or more glycosyltransferase genes into the cell by any of the methods described herein; and (d) introducing an expression vector comprising a transgene, the transgene comprising a sequence encoding a protein of interest, into the cell, thereby producing the recombinant protein. In certain embodiments, the protein of interest comprises, for example, erythropoietin or an antibody, for example, a monoclonal antibody.

Another aspect of the disclosure encompasses a method for producing a recombinant protein with a more homogeneous and/or human-like and/or novel and/or functionally beneficial glycosylation pattern. The method comprises expressing the protein in a mammalian cell line according to the present invention. In a specific embodiment, the cell line is a Chinese hamster ovary (CHO) cell line. In one embodiment, the cell line comprises inactivated chromosomal sequences encoding any endogenous glycosyltransferases. In one embodiment, the inactivated chromosomal sequences encoding any endogenous glycosyltransferases are monoallelic, and the cell line produces a reduced amount of said glycosyltransferases. In another embodiment, the inactivated chromosomal sequences encoding any endogenous glycosyltransferases are biallelic, and the cell line does not produce said measurable glycosyltransferases. In another embodiment, the recombinant protein has a more homogeneous and/or human-like and/or novel and/or functionally beneficial glycosylation pattern. In one embodiment, the recombinant protein has at least one improved property relative to a similar recombinant protein produced by a comparable cell line not deficient in said endogenous glycosyltransferases, for example, reduced immunogenicity, increased bioavailability, increased efficacy, increased stability, increased solubility, improved half-life, improved clearance, improved pharmacokinetics, and combinations thereof. The recombinant protein can be any protein, including a therapeutic protein. Exemplary proteins include those selected from, but not limited to, an antibody, an antibody fragment, a growth factor, a cytokine, a hormone, a lysosomal enzyme, a coagulation factor, a gonadotropin, and a functional fragment or variants thereof.

The disclosure may also be used to identify target genes for modification and use this knowledge to glycoengineer an existing mammalian cell line previously transfected with DNA encoding the protein of interest.

In any of the cells and methods described herein, the cell or cell line can be a COS, CHO (e.g., CHO-S, CHO-K1, CHO-DG44, CHO-DUXB11, CHOK1SV), VERO, MDCK, WI38, V79, B14AF28-G3, BHK, HaK, NS0, SP2/0-Ag14, HeLa, HEK293 (e.g., HEK293-F, HEK293-H, HEK293-T), and PERC6.

The cell lines described herein can also be used to produce other N-glycoproteins, including but not limited to, for example, α1-antitrypsin, gonadotropins, lysosomal enzyme proteins (e.g., glucocerebrosidase, α-galactosidase, α-glucosidase, sulfatases, glucuronidase, iduronidase). The known human glycosyltransferase genes are assembled into homologous gene families in the CAZy database and these families are further assigned to different glycosylation pathways in Hansen et al. (Hansen, Lind-Thomsen et al. 2014). TABLE 1 lists all human glycosyltransferase genes in the CAZy GT families with NCBI gene IDs and confirmed or putative role assignments in the biosynthesis of different mammalian glycoconjugates (N-glycans, O-GalNAc, O-GlcNAc, O-Glc, O-Gal, O-Fuc, O-Xyl, O-Man, C-Man, glycosphingolipids, hyaluronan, and GPI anchors). FIGURE 1 further graphically depicts the confirmed and putative roles of human CAZy GT families in the biosynthesis of different glycoconjugates.

Table 1. Human GTf genes

Family Identification Symbol® Description® CAZy(1) Genetic Location® on Map(5) GT32 53947 a1,4-galactosyltransferase 22q13.2 GT32 51146 a1,4-N-acetylglucosaminyltransferase 3p14.3 GT6 28 ABO Blood Type 9q34.2 GT33 56052 chitobiosyldiphosphodolichol p- 16p13.3 mannosyltransferase

GT59 84920 a1,2-glucosyltransferase 12p11.1 GT59 144245 a1,2-glucosyltransferase 12q 12 GT4 440138 a1,2-mannosyltransferase 13t 14.2 GT22 79087 a1,6-mannosyltransferase 22t13.33 GT1 79868 UDP-N subunit- Xq23

acetylglucosaminyltransferase

GT1 199857 UDP-N-1 subunit p21.3

acetylglucosaminyltransferase

GT33 200810 similar to achitobiosyldiphosphodolichol p- 3q21.2

mannosyltransferase -GT33 644974 chitobiosyldiphosphodolichol-like p- 3q22.1

mannosyltransferase-2

Family Identification Symbol Description® Genetic CAZy<1> location on map® GT4 85365ALG2a1,3/1,6-mannosyltransferase 9q22.33 GT58 10195 ALG3a1,3-mannosyltransferase 3q27.1

GT2 29880ALG5dolicyl-phosphate p-glucosyltransferase 13q13.3 GT57 29929Alg6a1,3-glucosyltransferase 1p31.3 GT57 79053ALG8 a1,3-glucosyltransferase 11q14.1

GT22 79796ALG9a1,2-mannosyltransferase 11q23 GT31 8706B3GALNT1p1,3-N-acetylgalactosaminyltransferase 1 3q25 GT31 148789B3GALNT2p1,3-N-acetylgalactosaminyltransferase 2 1q42.3 GT31 8708B3GALT1UDP-Gal:pGlcNAc p 1,3- 2q24.3 galactosyltransferase, polypeptide 1

GT31 8707B3GALT2UDP-Gal:pGlcNAc p1,3- 1q31

galactosyltransferase, polypeptide 2

GT31 8705B3GALT4UDP-Gal:pGlcNAc p1,3-6p21.3 galactosyltransferase, polypeptide 4

GT31 10317B3GALT5UDP-Gal:pGlcNAc p1,3-21q22.3 galactosyltransferase, polypeptide 5

GT31 126792B3GALT6UDP-Gal:pGal-p1,3- 1p36.33 galactosyltransferase polypeptide 6

GT31 145173B3GALTLp1,3-galactosyltransferase-like 13q12.3 GT43 27087B3GAT1p1,3-glucuronyltransferase 1 11q25 GT43 135152B3GAT2p1,3-glucuronyltransferase 2 6q13 GT43 26229B3GAT3p1,3-glucuronyltransferase 3 11q12.3 GT49 11041B3GNT1UDP-GlcNAc:pGal p1,3-N- 11q13.2 acetylglucosaminyltransferase 1

GT31 10678B3GNT2UDP-GlcNAc:pGal p1,3-N- 2p15

acetylglucosaminyltransferase 2

GT31 10331B3GNT3UDP-GlcNAc:pGal p1,3-N- 19p13.1 acetylglucosaminyltransferase 3

GT31 79369B3GNT4UDP-GlcNAc:pGal p1,3-N- 12q24 acetylglucosaminyltransferase 4

GT31 84002B3GNT5UDP-GlcNAc:pGal p1,3-N- 3q28

acetylglucosaminyltransferase 5

GT31 192134B3GNT6UDP-GlcNAc:pGal p1,3-N- 11q13.4 acetylglucosaminyltransferase 6

GT31 93010B3GNT7UDP-GlcNAc:pGal p1,3-N- 2q37.1 acetylglucosaminyltransferase 7

GT31 374907B3GNT8UDP-GlcNAc:pGal p1,3-N- 19q13.2 acetylglucosaminyltransferase 8

GT31 84752B3GNT9UDP-GlcNAc:pGal p1,3-N- 16q22.1 acetylglucosaminyltransferase 9

GT2 146712B3GNTL1 UDP-GlcNAc-like:pGal p1,3-N- 17q25.3 acetylglucosaminyltransferase 1

GT12 2583B4GALNT1p1,4-N-acetyl-galactosaminyltransferase 12q13.3

1

GT12 124872B4GALNT2p1,4-N-acetyl-galactosaminyltransferase 17q21.32

2

GT7 283358B4GALNT3p1,4-N-acetyl-galactosaminyltransferase 12p13.33

3

GT7 338707B4GALNT4p1,4-N-acetyl-galactosaminyltransferase 11p15.5

4

GT7 2683B4GALT1UDP-Gal:pGlcNAc p1,4- 9p13

galactosyltransferase, polypeptide 1

GT7 8704B4GALT2UDP-Gal:pGlcNAc p1,4- 1p34-p33 galactosyltransferase, polypeptide 2

GT7 8703B4GALT3UDP-Gal:pGlcNAc p1,4- 1q21-q23 galactosyltransferase, polypeptide 3

GT7 8702B4GALT4UDP-Gal:pGlcNAc p1,4- 3q13.3 galactosyltransferase, polypeptide 4

Family Identification Symbol® Description® Genetic CAZy(1) location® on map® GT7 9334B4GALT5UDP-Gal:pGlcNAc (31.4- 20q13.1-galactosyltransferase, polypeptide 5 q13.2 GT7 9331B4GALT6UDP-Gal:pGlcNAc 31.4- 18q 11 galactosyltransferase, polypeptide 6

GT7 11285B4GALT7xylosylprotein 31,4-galactosyltransferase, 5q35.2- polypeptide 7 q35.3 GT31 56913C1GALT1core 1 synthase, galactosyltransferase 1 7p21.3 GT31 29071C1GALT1C1C1-specific chaperone 1GALT1 Xq24 GT25 51148CERCAMbrain endothelial cell adhesion molecule 9q34.11

GT7/31 79586CHPFchondroitin polymerizing factor 2q35 GT7/31 54480CHPF2chondroitin polymerizing factor 2 7q36.1 GT7/31 22856CHSY1chondroitin sulfate synthase 1 15q26.3 GT7/31 337876CHSY3chondroitin sulfate synthase 3 5q23.3 GT25 79709COLGALT1collagen 3(1-O)galactosyltransferase 1 19p13.11 GT25 23127COLGALT2collagen 3(1-O)galactosyltransferase 2 1q25 GT7 55790CSGALNACT1chondroitin sulfate N- 8p21.3 acetylgalactosaminyltransferase 1

GT7 55454CSGALNACT2 chondroitin sulfate N- 10q 11.21 acetylgalactosaminyltransferase 2

GT2 8813DPM1 20q13.13 dolycylphosphate mannosyltransferase polypeptide 1

GTnc 23333DPY19L1dpy-19-like 1 (C. elegans) 7p14.3-p14.2 GTnc 283417DPY19L2dpy-192-like (C. elegans) 12q14.2 GTnc 147991DPY19L3dpy-193-like (C. elegans) 19q13.11 GTnc 286148DPY19L4dpy-194-like (C. elegans) 8q22.1 GT61 285203EOGTN-3p14.1 EGF domain-specific O-linked acetylglucosamine transferase

GT47/64 2131Ext1exostosin glycosyltransferase 1 8q24.11 GT47/64 2132EXT2exostosin glycosyltransferase 2 11 p12-p11 GT47/64 2134EXTL1glycosyltransferase 1 exostosin-like 1p36.1

GT64 2135Extl2glycosyltransferase 2 similar to 1p21

exostosin

GT47/64 2137EXTL3glycosyltransferase 3 similar to 8p21

exostosin

GTnc 79147FKRPfukutin-related protein 19q13.32 GTnc 2218FKTNFukutin 9q31-q33 GT11 2523FUT1fucosyltransferase 1, blood group H 19q13.3 GT10 84750FUT10fucosyltransferase 10, a1.3 8p12

fucosyltransferase

GT10 170384FUT11fucosyltransferase 11, a1,3 10q22.2 fucosyltransferase

GT11 2524FUT2 fucosyltransferase 2 secretory state 19q13.3 includes

GT10 2525FUT3fucosyltransferase 3, blood type 19p13.3 Lewis

GT10 2526FUT4fucosyltransferase 4, a1,3 11q21 fucosyltransferase, myeloid specific

GT10 2527FUT5fucosyltransferase 5, a1,3 19p13.3 fucosyltransferase

GT10 2528FUT6fucosyltransferase 6, a1,3 19p13.3 fucosyltransferase

GT10 2529FUT7fucosyltransferase 7, a1,3 9q34.3

fucosyltransferase

Family Identification Symbol® Description® Genetic CAZy(1) location® on map® GT23 2530 fucosyltransferase 8, a1,6 14q24.3 fucosyltransferase

GT10 10690 fucosyltransferase 9, a1,3 6q16

fucosyltransferase

GT27 2589 polypeptide N-18q 12.1 acetylgalactosaminyltransferase 1

GT27 55568 polypeptide N- 5q33.2 acetylgalactosaminyltransferase 10

GT27 63917 polypeptide N-7q36.1 acetylgalactosaminyltransferase 11

GT27 79695 polypeptide N- 9q22.33 acetylgalactosaminyltransferase 12

GT27 114805 polypeptide N-2q24.1 acetylgalactosaminyltransferase 13

GT27 79623 polypeptide N-2p23.1 acetylgalactosaminyltransferase 14

GT27 117248 polypeptide N-3p25.1 acetylgalactosaminyltransferase 15

GT27 57452 N-14q24.1 acetylgalactosaminyltransferase 16 polypeptide

GT27 374378 polypeptide N-11p15.3 acetylgalactosaminyltransferase 18

GT27 2590 polypeptide N- 1q41-q42 acetylgalactosaminyltransferase 2

GT27 2591 N-2q24-q31 acetylgalactosaminyltransferase 3 polypeptide

GT27 8693 polypeptide N- 12q21.33 acetylgalactosaminyltransferase 4

GT27 11227 polypeptide N-2q24.1 acetylgalactosaminyltransferase 5

GT27 11226 N-12q13 polypeptide acetylgalactosaminyltransferase 6

GT27 51809 polypeptide N-4q31.1 acetylgalactosaminyltransferase 7

GT27 26290 polypeptide N-12p13.3 acetylgalactosaminyltransferase 8

GT27 50614 polypeptide N- 12q24.33 acetylgalactosaminyltransferase 9

GT27 168391 N-7q36.1 acetylgalactosaminyltransferase 5 polypeptide-like

GT27 442117 N-4q34.1 acetylgalactosaminyltransferase 6 polypeptide-like

GT6 26301 globoside a1,3-N-acetylgalactosaminyltransferase 1 9q34.13 q34.3 GT14 2650 glucosaminyl (N-acetyl) transferase 1, 9q13

core 2

GT14 2651 glucosaminyl (N-acetyl) transferase 2, 6p24.2 branching enzyme I

GT14 9245 glucosaminyl (N-acetyl) transferase 3, type 15q21.3 mucin

GT14 51301 glucosaminyl (N-acetyl) transferase 4, 5q12

core 2

GT14 644378 glucosaminyl (N-acetyl) transferase 6 6p24.2 GT14 140687 glucosaminyl (N-acetyl) transferase family member 7 20q13.2

GT6 2681 glycoprotein, pseudogene of a-9q33.2 galactosyltransferase 1

GT4 144423 domain 1 containing 12q24.33 glycosyltransferase 1

GT6 360203 domain 1 containing 9q34.3

glycosyltransferase 6

Family Identification Symbol® Description® Genetic CAZy(1) location® on map® GT8 55830GLT8D1domain 1 containing 3p21.1 glycosyltransferase 8

GT8 83468GLT8D2domain 2 containing 12q glycosyltransferase 8

GT4 79712GTDC1 domain 1 similar containing 2q22.3

glycosyltransferase

GT8 283464GXYLT1glucoside xylosyltransferase 1 12q 12 GT8 727936GXYLT2xyloside xylosyltransferase 2 3p13 GT8 2992GYG1glycogenin 1 3q24-q25.1 GT8 8908GYG2glycogenin 2 Xp22.3 GT8/49 120071GYLTL1Bglycosyltransferase-like 1B, 11 p11.2

LARGE2

GT3 2997GYS1glycogen synthase 1 (muscle) 19q13.3 GT3 2998GYS2glycogen synthase 2 (liver) 12p12.2 GT2 3036HAS1hyaluronan synthase 1 19q13.4 GT2 3037HAS2hyaluronan synthase 2 8q24.12 GT2 3038HAS3hyaluronan synthase 3 16q22.1 GT90 79070KDELC1KDEL (Lys-Asp-Glu-Leu)-containing 1 13q33 glucosyltransferase-like

GT8/49 9215 LARGE p1,3-xylosyltransferase 22q12.3 GT31 3955LFNGO-fucosylpeptide 3-pN- 7p22.2 acetylglucosaminyltransferase

GT31 4242MFNGO-fucosylpeptide 3-pN- 22q12 acetylglucosaminyltransferase

GT13 4245MGAT1mannosyl a1,3-glycoprotein p1,2N- 5q35

acetylglucosaminyltransferase

GT16 4247MGAT2mannosyl a1,6-glycoprotein p1,2N- 14q21 acetylglucosaminyltransferase

GT17 4248MGAT3mannosyl a1,4-glycoprotein p1,4N- 22q13.1 acetylglucosaminyltransferase

GT54 11320MGAT4Amanosyl a1,3-glycoprotein p1,4N- 2q12

acetylglucosaminyltransferase

GT54 11282MGAT4Bmannosyl a1,3-glycoprotein p1,4N- 5q35

acetylglucosaminyltransferase

GT54 25834MGAT4Cmannosyl a1,3-glycoprotein p1,4N- 12q21 acetylglucosaminyltransferase

GT18 4249MGAT5mannosyl a1,6-glycoprotein p1,6N-2q21.3 acetylglucosaminyltransferase

GT18 146664MGAT5Bmannosyl a1,6-glycoprotein p1,6N- 17q25.2 acetylglucosaminyltransferase, isoenzyme

B

GT41 8473OGTN-O-linked acetylglucosamine (GlcNAc) Xq13

transferase

GT4 5277PIGAbiosynthesis of glycan anchors from Xp22.1 phosphatidylinositol, class A

GT22 9488PIGB15q21.3 phosphatidylinositol glycan anchor biosynthesis, class B

GT50 93183PIGM 1q23.2 phosphatidylinositol glycan anchor biosynthesis, class M

GT76 55650PIGVGlycan anchor biosynthesis of 1 p36.11 phosphatidylinositol, class V

GT22 80235PIGZ 3q29 glycan anchor biosynthesis

phosphatidylinositol, class Z

GTnc 8985PLOD3procollagen-lysine, 2-oxoglutarate 5- 7q22

dioxygenase 3

GT65 23509POFUT1O-fucosyltransferase 1 20q11 GT68 23275 POFUT2O-fucosyltransferase 2 21q22.3 Family ID Symbol® Description® Genetic CAZy(1) location® on map® GT90 56983POGLUT1O-glucosyltransferase 1 3q13.33 GT13 55624POMGNT1O-mannose-linked p1,2-N-1p34.1 acetylglucosaminyltransferase 1

GT61 84892POMGNT2O-linked-mannose p1,4-N- 3p22.1 acetylglucosaminyltransferase 2

GT39 10585POMT1O-mannosyltransferase 1 9q34.1 GT39 29954POMT2O-mannosyltransferase 2 14q24 GT35 5834PYGBphosphorylase, glycogen; brain 20p11.21 GT35 5836PYGLphosphorylase, glycogen, liver 14q21-q22 GT35 5837PYGMphosphorylase, glycogen, muscle 11q12-q13.2 GT31 5986RFNGO-fucosylpeptide 3-pN- 17q25 acetylglucosaminyltransferase

GT29 6482ST3GAL1p-galactoside a-2,3-sialyltransferase 1 8q24.22 GT29 6483ST3GAL2p-galactoside a-2,3-sialyltransferase 2 16q22.1 GT29 6487ST3GAL3p-galactoside a-2,3-sialyltransferase 3 1p34.1 GT29 6484ST3GAL4p-galactoside a-2,3-sialyltransferase 4 11q24.2 GT29 8869ST3GAL5p-galactoside a-2,3-sialyltransferase 5 2p11.2 GT29 10402ST3GAL6p-galactoside a-2,3-sialyltransferase 6 3q 12.1 GT29 6480ST6GAL1p-galactosamide a-2,6-sialyltransferase 1 3q27-q28 GT29 84620ST6GAL2p-galactosamide a-2,6-sialyltransferase 2 2q11.2-q12.1 GT29 55808ST6GALNAC1a-N-acetyl-neuraminyl-2,3-p-galactosyl- 17q25.1

1,3-N-acetylgalactosaminide a-2,6-sialyltransferase 1

GT29 10610ST6GALNAC2a-N-acetyl-neuraminyl-2,3-p-galactosyl- 17q25.1

1,3-N-acetylgalactosaminide a-2,6-sialyltransferase 2

GT29 256435ST6GALNAC3a-N-acetyl-neuraminyl-2,3-p-galactosyl- 1p31.1

1,3-N-acetylgalactosaminide a-2,6-sialyltransferase 3

GT29 27090ST6GALNAC4a-N-acetyl-neuraminyl-2,3-p-galactosyl- 9q34

1,3-N-acetylgalactosaminide a-2,6-sialyltransferase 4

GT29 81849ST6GALNAC5a-N-acetyl-neuraminyl-2,3-p-galactosyl- 1p31.1

1,3-N-acetylgalactosaminide a-2,6-sialyltransferase 5

GT29 30815ST6GALNAC6a-N-acetyl-neuraminyl-2,3-p-galactosyl- 9q34.11

1,3-N-acetylgalactosaminide a-2,6-sialyltransferase 6

GT29 6489ST8SIA1o-N-acetyl-neuraminide a-2,8- 12p12.1-sialyltransferase 1 p11.2 GT29 8128ST8SIA2o-N-acetyl-neuraminide a-2,8- 15q26 sialyltransferase 2

GT29 51046ST8SIA3o-N-acetyl-neuraminide a-2,8- 18q21.31 sialyltransferase 3

GT29 7903ST8SIA4a-N-acetyl-neuraminide a-2,8- 5q21

sialyltransferase 4

GT29 29906ST8SIA5o-N-acetyl-neuraminide a-2,8- 18q21.1 sialyltransferase 5

GT29 338596ST8SIA6o-N-acetyl-neuraminide a-2,8- 10p12.33 sialyltransferase 6

GT66 3703STT3A subunit of the 11q23.3 oligosaccharyltransferase complex (catalytic)

GT66 201595STT3B subunit of the 3p23 complex

oligosaccharyltransferase (catalytic)

GTnc 10329 TMEM5 transmembrane protein 5 12q14.2 Family Identification Symbol Description® Genetic CAZy(1) location® on map® GT21 7357ST8SIAUDP-glucose ceramide 9q31

glucosyltransferase

GT24 56886UGGT1 UDP-glucose glycoprotein 2q14.3

glucosyltransferase 1

GT24 55757UGGT2UDP-glucose glycoprotein 13q32.1 glucosyltransferase 2

GT1 54658UGT1A1UDP glucuronosyltransferase 2q37 family

1, polypeptide A1

GT1 54575UGT1A10UDP glucuronosyltransferase 1, 2q37 family

A10 polypeptide

GT1 54659UGT1A3UDP family glucuronosyltransferase 1, 2q37

A3 polypeptide

GT1 54657UGT1A4UDP family glucuronosyltransferase 1, 2q37

A4 polypeptide

GT1 54579UGT1A5UDP glucuronosyltransferase 1, 2q37 family

A5 polypeptide

GT1 54578UGT1A6UDP glucuronosyltransferase 1, 2q37 family

A6 polypeptide

GT1 54577UGT1A7UDP glucuronosyltransferase 1, 2q37 family

A7 polypeptide

GT1 54576UGT1A8UDP glucuronosyltransferase 1, 2q37 family

A8 polypeptide

GT1 54600UGT1A9UDP family glucuronosyltransferase 1, 2q37

A9 polypeptide

GT1 10941UGT2A1UDP glucuronosyltransferase 2 family, 4q13

A1 polypeptide

GT1 79799UGT2A3UDP glucuronosyltransferase 2 family, 4q13.2

A3 polypeptide

GT1 7365UGT2B10UDP glucuronosyltransferase 2 family, 4q13.2

B10 polypeptide

GT1 10720UGT2B11UDP glucuronosyltransferase 2 family, 4q13.2

B11 polypeptide

GT1 7366UGT2B15UDP glucuronosyltransferase 2 family, 4q13

B15 polypeptide

GT1 7367UGT2B17UDP glucuronosyltransferase 2 family, 4q13

B17 polypeptide

GT1 54490UGT2B28UDP glucuronosyltransferase 2 family, 4q13.2

B28 polypeptide

GT1 7363UGT2B4UDP glucuronosyltransferase 2 family, 4q13

polypeptide B4

GT1 7364UGT2B7UDP glucuronosyltransferase 2, 4q 13 family

B7 polypeptide

GT1 133688UGT3A1UDP glycosyltransferase 3 family, 5p13.2

A1 polypeptide

GT1 167127UGT3A2UDP glycosyltransferase 3 family, 5p13.2

A2 polypeptide

GT1 7368UGT8UDP glycosyltransferase 8 4q26 GT27 64409WBSCR17Chromosomal region 17 of 7q11.23 Williams-Beuren syndrome

GT8 152002XXYLT1 xyloside xylosyltransferase 1 3q29

GT14 64131 XYLT1xylosyltransferase I 16p12.3

GT14 64132XYLT2xylosyltransferase II 17q21.33 (1) GT classification system (Lombard et al. (2013). Nucí Acid Res 42:D1P:D490-D495)(2) gene symbol(3) GenBank gene ID(4) UniProt protein name(5) Human chromosomal position (hg19)_____

Most of the corresponding orthologous CHO glycosyltransferase genes were previously assigned in connection with the recent sequencing of the CHO genome (Xu, Nagarajan et al. 2011), but some genes were incorrectly assigned or omitted. The current set of orthologous glycosyltransferase genes in humans and CHO are listed in Table 2.

Table 2. GTfs genes of the Chinese hamster (Cricetulus griseus) Family CAZy(1) Symbol(2) Gene ID(3) Description*4)

CAZy Family(1) Symbol(2) Identification Description*4*

genetics*3*

GT1Alg13100754023 UDP-N-acetylglucosaminyltransferase subunit

GT1Alg14100773644 UDP-N-acetylglucosaminyltransferase subunit

GT1U g tla l100755423 UDP-glucuronosyltransferase 1-6

GT1Ugt2a1100762963 UDP glucuronosyltransferase family 2, polypeptide A1, complex locus

GT1Ugt2a3100762673 UDP-glucuronosyltransferase 2A3

GT2Alg5100769679 ALG5, dolicyl-phosphate p-glucosyltransferase GT2B3gntl1100751193 similar to UDP-GlcNAc:pGal p-1,3-N-acetylglucosaminyltransferase 1

GT2Dpm1100689420 dolichyl-phosphate mannosyltransferase polypeptide 1, catalytic subunit

GT2Has2100751055 hyaluronan synthase 2

GT2Has2100751055 hyaluronan synthase 2

GT2Has3100757895 hyaluronan synthase 3

GT3Gys1100769788 glycogen synthase 1 (muscle)

GT3Gys2100770628 glycogen synthase 2 (liver)

GT4Alg11100771009 a-1,2-mannosyltransferase

GT4Alg2100768412 a-1,3/1,6-mannosyltransferase

GT4Glt1d1100761757 domain 1 containing glycosyltransferase 1 GT4Gtdc1100752161 domain 1 containing glycosyltransferase similar

GT4Piga100773712 biosynthesis of phosphatidylinositol glycan anchors, class A

GT6Abo100772592 ABO blood group

GT6Gbgt1100771256 globoside 1 p 1,3-N-acetylgalactosaminyltransferase

GT7B4galnt3100756528 p-1,4-N-acetylgalactosaminyltransferase 3 GT7B4galnt4100758404 p-1,4-N-acetylgalactosaminyltransferase 4 GT7B4galt1100689430 UDP-Gal:pGlcNAc p1,4-galactosyltransferase, polypeptide 1 GT7B4galt2100689434 UDP-Gal:pGlcNAc p1,4-galactosyltransferase, polypeptide 2 GT7B4galt3100689346 UDP-Gal:pGlcNAc p1,4-galactosyltransferase, polypeptide 3 GT7B4galnt4100689435 UDP-Gal:pGlcNAc p1,4-galactosyltransferase, polypeptide 4 GT7B4galt5100689347 UDP-Gal:pGlcNAc p1,4-galactosyltransferase, polypeptide 5 GT7B4galt6100689438 UDP-Gal:pGlcNAc p1,4-galactosyltransferase, polypeptide 6 GT7B4galt7100769652 xylosylprotein p1,4-galactosyltransferase, polypeptide 7

GT7/GT3 1Chpf100765856 chondroitin polymerizing factor GT7/GT3 1Chsy1100770803 chondroitin sulfate synthase 1

GT7/GT3 1Chsy3100767985 chondroitin sulfate synthase 3

GT7Csgalnact1100754969 chondroitin sulfate N-acetylgalactosaminyltransferase 1

GT7Csgalnact2100770830 chondroitin sulfate N-acetylgalactosaminyltransferase 2

GT8Glt8d1100762757 domain 1 containing glycosyltransferase 8 GT8Glt8d2100772458 domain 2 containing glycosyltransferase 8 GT8Gxylt1100760543 glycoside xylosyltransferase 1

GT8Gxylt2100758512 xyloside xylosyltransferase 2

GT8Gyg1100751671 glycogenin 1

GT8/49Gyltl1b100750932 glycosyltransferase-like 1B GT8/49Grande100765249 laglycosyltransferase-like

GT8Xxylt1100760146 xyloside xylosyltransferase 1

CAZy Family(1) Symbol Identification Description®

genetics®

GT10Fut10100760260 fucosyltransferase 10 (a(1,3)fucosyltransferase)

GT10Futí 1100758336 fucosyltransferase 11 (a (1,3) fucosyltransferase)

GT10Fut4100754814 fucosyltransferase 4 (a(1,3) fucosyltransferase, myeloid specific) GT10Fut6a100689084 a1,3 fucosyltransferase 6A

GT10Fut6b100689083 a1,3 fucosyltransferase 6B

GT10Fut7100772214 a1,3 fucosyltransferase

GT10Fut9100689036 a1,3 fucosyltransferase

GT11Futí100757047 galactoside 2o-L-fucosyltransferase, blood group H

GT11Fut2100751185 fucosyltransferase 2

GT12B4gaIn100764682 p1,4-N-acetyl-galactosaminyltransferase 1 GT12B4gaIn100760696 p1,4-N-acetyl-galactosaminyltransferase 2 GT13Mgatí100682529 mannosyl (a1,3)-glycoprotein p1,2N-acetylglucosaminyltransferase

GT13Pomgn100772511 O-linked protein mannose N-acetylglucosaminyltransferase 1 (p1,2-) GT14Gentí100767124 glucosaminyl (N-acetyl) transferase 1, nucleus

2

GT14Gen t3100774815 glucosaminyl (N-acetyl) transferase 3, mucin type

GT14Gen t4100757239 glucosaminyl (N-acetyl) transferase 4, nucleus

2

GT14Gcnt7100760777 member 7 of the glucosaminyl (N-acetyl) transferase family

GT14Xyltí100756878 xylosyltransferase I

GT14Xylt2100759604 xylosyltransferase II

GT16Mgat2100753385 mannosyl (α1,6)-glycoprotein p-1,2-N-acetylglucosaminyltransferase

GT17Mgat3100689076 mannosyl(p1,4)-glycoprotein p-1,4-N-acetylglucosaminyltransferase

GT18Mgat5100760162 mannosyl(p 1,6)-glycoprotein p-1,6-N-acetylglucosaminyltransferase

GT18Mgat5b100771275 mannosyl(a1,6)-glycoprotein p-1,6-N-acetylglucosaminyltransferase

GT21Ugcg100689432 UDP-glucose ceramide glucosyltransferase GT22Algí2100770096 a-1,6-mannosyltransferase

GT22Alg9100755062 a-1,2-mannosyltransferase

GT22Pigb100768002 Biosynthesis of phosphatidylinositol glycan anchors, class B

GT22Pigz100750622 biosynthesis of phosphatidylinositol glycan anchors, class Z

GT23Fut8100751648 a-1,6-mannosyltransferase)

GT24Uggtí100773968 UDP-glucose glycoprotein glucosyltransferase 1

GT24Uggt2100762273 UDP-glucose glycoprotein glucosyltransferase 2

GT25Cerca100765284 brain endothelial cell adhesion molecule

GT25Colgalt100774081 collagen p(1-O)galactosyltransferase 1 GT25Colgalt100764465 collagen p(1-O)galactosyltransferase 2 GT27Galntí100763868 polypeptide N-acetylgalactosaminyltransferase 1 GT27Galntí100768523 polypeptide N-acetylgalactosaminyltransferase 10 GT27Galntí100758359 polypeptide N-acetylgalactosaminyltransferase 11 GT27Galntí100772146 polypeptide N-acetylgalactosaminyltransferase 12 CAZy Family(1) Symbol*2) Identification Description*4)

genetics*3)

GT27Galnt13100768831 polypeptide N-acetylgalactosaminyltransferase 13 GT27Galnt14100759781 polypeptide N-acetylgalactosaminyltransferase 14 GT27Galnt15100766936 polypeptide N-acetylgalactosaminyltransferase 15 GT27Galnt16100764829 polypeptide N-acetylgalactosaminyltransferase 16 GT27Galnt18100767393 polypeptide N-acetylgalactosaminyltransferase 18 GT27Galnt2100767525 polypeptide N-acetylgalactosaminyltransferase 2 GT27Galnt3100753226 polypeptide N-acetylgalactosaminyltransferase 3 GT27Galnt4100765247 polypeptide N-acetylgalactosaminyltransferase 4 GT27Galnt5100757219 polypeptide N-acetylgalactosaminyltransferase 5 GT27Galnt6100751126 polypeptide N-acetylgalactosaminyltransferase 6 GT27Galnt6100751126 polypeptide N-acetylgalactosaminyltransferase 6 GT27Galnt7100762043 polypeptide N-acetylgalactosaminyltransferase 7 GT27Galnt8100764127 polypeptide N-acetylgalactosaminyltransferase 8 GT27Galnt9100773797 polypeptide N-acetylgalactosaminyltransferase 9 GT27Galntl5100767659 similar to polypeptide N-acetylgalactosaminyltransferase 5 GT27Galntl6100761752 N-acetylgalactosaminyltransferase 6-like polypeptide GT27Wbser17100750837 Williams-Beuren syndrome chromosome region 17

GT29St3gal1100754088 p-galactoside a-2,3-sialyltransferase 1 GT29St3gal1100754088 p-galactoside a-2,3-sialyltransferase 1 GT29St3gal2100767717 p-galactoside a-2,3-sialyltransferase 2 GT29St3gal2100767717 p-galactoside a-2,3-sialyltransferase 2 GT29St3gal3100689187 p-galactoside a-2,3-sialyltransferase 3 GT29St3gal4100689440 p-galactoside a-2,3-sialyltransferase 4 GT29St3gal5100754838 p-galactoside a-2,3-sialyltransferase 5 GT29St3gal6100771326 p-galactoside a-2,3-sialyltransferase 6 GT29St6gal1100689389 p-galactosamide a-2,6-sialyltransferase 1 GT29St6gal2100763756 p-galactosamide a-2,6-sialyltransferase 2 GT29St6galnac1100763224 o-N-acetylgalactosaminide a-2,6-sialyltransferase 1

GT29St6galnac3100762285 (a-N-acetyl-neuraminyl-2,3-beta-galactosyl-1.3)-N-acetylgalactosaminide a-2,6-sialyltransferase 3

GT29St6galnac4100759065 (a-N-acetyl-neuraminyl-2,3-beta-galactosyl-1.3)-N-acetylgalactosaminide a-2,6-sialyltransferase 4

GT29St6galnac5100759381 (a-N-acetyl-neuraminyl-2,3-beta-galactosyl-1.3)-N-acetylgalactosaminide a-2,6-sialyltransferase 5

GT29St6galnac6100757138 (a-N-acetyl-neuraminyl-2,3-beta-galactosyl-1.3)-N-acetylgalactosaminide a-2,6-sialyltransferase 6

GT29St8sia 1100768920 o-N-acetyl-neuraminide a-2,8-sialyltransferase 1

GT29St8sia2100759559 o-N-acetyl-neuraminide a-2,8- sialyltransferase 2

CAZy Family(1) Symbol*2) Identification Description*4)

genetics*3)

GT29St8sia3100764454 o-N-acetyl-neuraminide a-2,8-sialyltransferase 3

GT29St8sia4100689217 o-N-acetyl-neuraminide a-2,8-sialyltransferase 4

GT29St8sia5100750766 o-N-acetyl-neuraminide a-2,8-sialyltransferase 5

GT29St8sia6100774188 o-N-acetyl-neuraminide a-2,8-sialyltransferase 6

GT31B3galnt1100756438 3-1,3-N-acetylgalactosaminyltransferase 1

(globoside blood group)

GT31B3galnt2100768564 31,3-N-acetylgalactosaminyltransferase 2 GT31B3galt1100761765 UDP-Gal:pGlcNAc (31,3-galactosyltransferase, polypeptide 1 GT31B3galt2100766715 UDP-Gal:pGlcNAc 31,3-galactosyltransferase, polypeptide 2 GT31B3galt4100751197 UDP-Gal:pGlcNAc 31,3-galactosyltransferase, polypeptide 4 GT31B3galt5100755714 UDP-Gal:3GlcNAc 31,3-galactosyltransferase, polypeptide 5 GT31B3galt6100767598 UDP-Gal:3GlcNAc 31,3-galactosyltransferase, polypeptide 6 GT31B3galtl100759864 Similar to 31,3-galactosyltransferase GT31B3gnt2100766691 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 2

GT31B3gnt3100773213 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 3

GT31B3gnt4100769077 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 4

GT31B3gnt5100757919 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 5

GT31B3gnt6100768656 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 6

GT31B3gnt7100760456 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 7

GT31B3gnt8103160327 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 8

GT31B3gnt9103159649 UDP-GIcNAc:3Gal 3-1,3-N-acetylglucosaminyltransferase 9

GT31C lga ltl100761169 glycoprotein-N-acetylgalactosamine 3-galactosyltransferase, 1

GT31C1galt1c1100751243 C1GALT1-specific chaperone 1 GT31Lfng100762397 O-fucosylpeptide 3-3-N-acetylglucosaminyltransferase

GT31Mfng100762875 O-fucosylpeptide 3-3N-acetylglucosaminyltransferase

GT31Rfng100771257 O-fucosylpeptide 3-3N-acetylglucosaminyltransferase

GT32A4galt100770462 a-1,4-galactosyltransferase

GT32A4gnt100771969 a-1,4-N-acetylglucosaminyltransferase GT33Alg1100773731 chitobiosyldiphosphodolichol 3-mannosyltransferase

GT35Pygb100769186 phosphorylase, glycogen; brain

GT35Pygm100757350 phosphorylase, glycogen, muscle

GT39Pomt1100755033 protein-O-mannosyltransferase 1 GT39Pomt2100764752 protein-O-mannosyltransferase 2 GT41Ogt100768670 O-linked N-acetylglucosamine (GlcNAc) transferase

GT43B3gat1100750701 3-1,3-glucuronyltransferase 1

(glucuronosyltransferase P)

GT43B3gat2100756696 3-1,3-glucuronyltransferase 2

(glucuronosyltransferase S)_______________ CAZy Family(1) Symbol(2) Identification Description®

genetics®

GT43B3gat3100689419 (3-1,3-glucuronyltransferase 3

(glucuronosyltransferase I)

GT47/64Ext1100689334 exostosin glycosyltransferase 1 GT47/64Ext2100751585 exostosin glycosyltransferase 2 GT47/64Extll100770000 exostosin-like glycosyltransferase 1 GT47/64Extl3100751999 exostosin-like glycosyltransferase 3 to exostosin GT49B3gnt1100762253 UDP-GIcNAc:pGal p-1,3-N-acetylglucosaminyltransferase 1

GT50Pigm100764842 biosynthesis of phosphatidylinositol glycan anchors, class M

GT54Mgat4a100766200 mannosyl(a1,3)-glycoprotein p-1,4-N-acetylglucosaminyltransferase

GT54Mgat4b100768637 mannosyl(a1,3)-glycoprotein p-1,4-N-acetylglucosaminyltransferase

GT54Mgat4c100760589 mannosyl (a-1,3)-glycoprotein (3-1,4-N-acetylglucosaminyltransferase

GT57Alg6100753783 a-1,3-glucosyltransferase

GT57Alg8100766150 a-1,3-glucosyltransferase

GT58Alg3100772003 a-1,3- mannosyltransferase

GT61Eogt100757071 EGF domain-specific O-linked N-acetylglucosamine (GlcNAc) transferase GT61Pomgnt2100770708 O-linked mannose protein N-acetylglucosaminyltransferase 2 (p1,4-) GT64Extl2100761218 exostosin-like glycosyltransferase 2 GT65Pofut1100753417 O-fucosyltransferase protein 1

GT66Stt3a100751391 subunit of the oligosaccharyltransferase complex (catalytic) GT66Stt3b100752084 subunit of the oligosaccharyltransferase complex (catalytic) GT68Pofut2100765129 protein O-fucosyltransferase 2

GT90Kdelc1100763723 KDEL (Lys-Asp-Glu-Leu) containing 1 GT90Poglut1100760325 protein O-glycosyltransferase 1

GTncDpy19/1100765722 similar to dpy-19 1 (C. elegans)

GTncDpy19/2100752353 similar to dpy-192 (C. elegans)

GTncDpy19/3100756435 similar to dpy-193 (C. elegans)

GTncDpy19l4100759810 similar to dpy-194 (C. elegans)

GTncFktn100761516 fukutin

GTncPlod3100768993 procollagen-lysine, 2-oxoglutarate 5-dioxygenase 3

GTncTmem5100757873 transmembrane protein 5

-Dpagt1100689054 dolichyl-phosphate (UDP-N-acetylglucosamine) N-acetylglucosaminephosphotransferase 1 (1) GT classification system (Lombard et al. (2013). Nucí Acid Res 42:D1 P:D490-D495)(2) gene symbol HGNC(3) GenBank gene ID(4) Protein name UniProt______________________________________

Similarly, the transcriptome data for a CHO-K1 clone reported in Xu et al. (Xu, Nagarajan et al. 2011) did not include several glycosyltransferase genes. Therefore, we performed RNA sequencing analysis of a panel of CHO lines, including those with ZFN-mediated knockout of glycosyltransferase genes, to assess the expression of all glycosyltransferase genes in CHO. TABLE 3 provides a list of the expression levels of all assigned CHO glycosyltransferase genes, and some qualitative differences in expression levels were found compared with those previously reported (Xu, Nagarajan et al. 2011).

Table 3. Comparison of GTfs expression levels from RNA_seq data for CHO-GS and CHO-K1

Depth of CHO-K1 Family CAZyGenes GTf (hGTfs from Table CHO- eq (Xu et al 1)* GS_FPKM ** (4) RNAs

2011 )**(4)

GT1Alg1313 na

l

l

l

l

l

* A total of 208 hGTfs are reduced to 200 more Dpagtl (a non-GTf gene) resulting in a total of 201 genes included in the RNA_seq data comparison. ** Expression levels are not directly comparable with the CHO-GS data. They are FRKP and for the CHO-K1 data (Xu et al. 2011), read depth is taken as a measure of expression. Na indicates genes not analyzed due to missing annotations. (1) The Ugtl and Ugt2a/b genes are highly heterogeneous between primates and rodents, and generally encode the largest and most heterogeneous set of glucuronyltransferase genes in rodent genomes. (2) Ugt1a1-10 represent a locus with 9 genes; these have been combined into one gene in the RNA_seq data analysis. (3) Dgagt1 is not included in the list of human GTfs. (4) Na - no analyzed____________________

To explore the functions of individual glycosyltransferase genes putatively involved in N-glycosylation, the present invention employed a ZFN-mediated knockout assay in a CHO-GS cell line (Sigma), and also in other cell lines. The present invention designed a knockout screen to dissect the in vivo functions of all genes encoding isozymes with the potential to regulate N-glycosylation branching, elongation, and end-capping expressed in CHO (Figure 14).

The present invention probed the effects of the knockout screen on N-glycosylation capacity by expressing recombinant human erythropoietin (EPO) in mutant CHO cell lines and analyzing the glycosylation of purified EPO by N-glycan release and profiling using MALDI-TOF. EPO was used as an indicator of N-glycosylation capacity because it is one of the best characterized N-glycoproteins produced in CHO, with three N-glycans with mainly tetraantennary structure, low level of poly-LacNAc, and a 2,3-sialic acid coating (Figure 15) (Sasaki, Ochi et al. 1988). EPO expressed in the unmodified CHO-GS production line grown in suspension in protein-free medium was glycosylated, essentially identical to previous reports of EPO produced in CHO-K1 (Figure 15). Furthermore, the glycosylation profile was essentially identical to that of the EPO therapeutic products produced in CHO and claimed in drug applications (Sasaki, Bothner et al. 1987).

The present invention utilized ZFNs, TALENs, and CRISPR/Cas9 to target and knockout 19 glycosyltransferase genes (Figure 14A) involved in N-glycan branching (Mgat1/2/3/4A/4B/4C/5/5B) (FIG 5, FIG 21B), galactosylation (B4galt1/2/3/4 (Figure 17), poly-LacNAc elongation (B3gnt1/2/8) (Figure 18), end-capping by sialylation (st3gal3/4/6) (Figure 19), and core fucosylation (fut8) (Figure 22).

N-glycan antenna state: The biosynthetic control of the N-glycan antenna state was first explored. MGAT1 and 2 each control the formation of one of the two branches (32 in biantennary N-glycans, while a possible partial functional redundancy is predicted for tri- and tetraantennary branch formation by MGAT4A/4B/4C (p4 branch) and MGAT5/5B (p6 branch), respectively. Only MGAT4B and 5 were found to be substantially expressed in CHO as assessed by RNAseq results (TABLE 3), however, surprisingly, targeting the Mgat4a gene had a clear albeit minor effect, and only targeting Mgat4a/4b completely abolished p4 branched tetraantennary N-glycans (Figure 16A-B).

In contrast, targeting Mgat5 abolished L-PHA lectin labeling of cells (Figure 20) and p6 branched tetra-antennary structures (Figure 5), consistent with studies of the Lec4 mutant (Patnaik and Stanley 2006; North, Huang et al. 2010). Furthermore, Mgat4a/4b/5 KO stacking produced nearly homogeneous biantennary N-glycans with a decreased amount of poly-LacNAc (Figure 16B), which is also found as a minor component in wild-type (WT) cells (North, Huang et al. 2010).

LacNAc control: CHO cells express all seven known β4-galactosyltransferases (TABLE 3), and our understanding of the in vivo functions of these isozymes is poor. Only β4galt1-β4 are expected to play roles in N-glycosylation and have been suggested to have preferences for different N-glycan branches. The present invention first examined individual KO of B4galt1-β4 in WT CHO-GS and confirmed that B4galt1 appeared to play the primary role in LacNAc formation using EPO as a reporter molecule (Figure 17). Therefore, the invention could only assess the contributions of the other isoforms in stacked combinations with β4galt1 KO. Probing the stacked combinations by immunocytology with an antibody against LacNAc surprisingly showed that only the stacked KO of B4galt1/3, and not B4galt1/2/4, appeared to abolish galactosylation (FIG 10). In agreement with this, the invention found that EPO expressed in B4galt1/3 but not B4galt1/2 stacked KO clones had a substantial (>90%) reduction in galactosylation (FIG 17).

The invention further tested stacked KO of individual B4galts in CHO-GS Mgat4a/4b/5 KO cell lines with homogeneous biantennary N-glycans, and found that KO of B4galt1 abolished immunoreactivity for LacNAc while KO of B4galt2, 3, and 4 had no substantial effects (Figure 21A-B). This suggested that the galactosyltransferase isoform encoded by B4galt1 is the major or only isozyme capable of transferring Gal to biantennary N-glycans, while the other isoforms also function with tri- and tetraantennary N-glycans. To test this further, the invention included a human therapeutic IgG as a recombinant expressed reporter molecule (Figure 22). Human IgGs typically have only one N-glycan in the conserved region at Asn297, and this site is usually only glycosylated with biantennary N-glycans with variable and heterogeneous degrees of galactosylation and sialylation. Some IgG molecules have additional N-glycan sites in the variable regions and these are usually glycosylated with more complex tri- and tetraantennary N-glycan structures with efficient galactosylation and sialic acid protection. Importantly, recombinant expression of an IgG in B4galt1 knockout CHO-GS resulted in homogeneous biantennary N-glycans without galactosylation (Figure 22). Since CHO normally only produce IgG with biantennary structures, it follows that only deletion of B4galt1 is required to achieve this. This demonstrates that the inherent heterogeneity found in the N-glycosylation of therapeutic monoclonal antibodies can be eliminated by using this modified CHO cell line.

Since CHOs produce a high level of a6Fuc on N-glycans and since elimination of this glycosylation can be achieved by FUT8 knockout, it is clear that knockout combinations of Mgat4a/4b/5, B4galt1, and Fut8, as well as stacked KO of B4galt1 and Fut8, will generate CHO clones producing homogeneous N-glycans without a6Fuc. These production cell lines will be important for producing therapeutic IgG antibodies with enhanced ADCC.

Poly-LacNAc control - Poly-LacNAc on N-glycans has consistently been found to be incomplete and heterogeneous (Fukuda, Sasaki et al. 1989; Takeuchi, Inoue et al. 1989; North, Huang et al. 2010). CHO cells generally produce low amounts of poly-LacNAc on N-glycans and mainly on tetra-antennary structures on the p6 branch controlled by MGAT5 (North, Huang et al. 2010). The biosynthesis of poly-LacNAc on N-glycans is poorly understood and three genes, B3gnt1, B3gnt2 and B3gnt8, are candidates (Narimatsu 2006). Using single-gene KO of B3gnt1, B3gnt2, and B3gnt8 with LEL lectin immunocytology, the current invention identified B3gnt2 as the key gene controlling LacNAc initiation in CHO cells (Figure 23), and EPO expressed in CHO with KO of B3gnt2 but not B3gnt1 lacked poly-LacNAc (Figure 18). Furthermore, KO of B3gnt2 in CHO with additional KO of Mgat4a/4b/5 resulted in EPO with homogeneous biantennary N-glycans without poly-LacNAc (Figure 18). It was surprising that B3GNT1 did not play a role, as the gene was originally identified by expression cloning as the poly-LacNAc synthase (Sasaki, Kurata-Miura et al. 1997), although this gene was later shown to also function in O-mannosylation (Bao, Kobayashi et al. 2009). Surprisingly, KO of B3gnt8 did not weaken LEL lectin immunocytogenesis. B3gnt8 was previously shown to form heterodimers in vitro with B3gnt2 and to activate B3gnt2 enzymatic activity.

Sialylation control: CHO cells exclusively produce a NeuAc coating on α2,3-linked N-glycans (Watson, Bhide et al. 1994). A few reports have suggested the presence of minor amounts of α2,6-linked NeuAc as well as NeuGc, and genes for the synthesis of both α2,6-linked and CMP-NeuGc are found in CHO cells but are not normally expressed (TABLE 3). However, one or both of these genes can be inactivated by the present invention and in previous reports (US Patent US20130004992). All six known st3gal1-6 sialyltransferases are expressed in CHO cells (TABLE 3) and the roles of each in N-glycan sialylation are poorly understood (Tsuji, Datta et al. 1996). The sialyltransferase genes st3gal3, st3gal4, and st3gal6 were first analyzed individually by immunocytology to assess LacNAc exposure, and it was found that only KO of st3gal4 resulted in substantial LacNAc exposure (Figure 24). Analysis of EPO expressed in single and stacked KO clones showed that loss of st3gal4 and 6 resulted in substantial exposure. In WT CHO, KO of st3gal4/6 resulted in EPO with heterogeneous tetraantennary N-glycans without sialylation (Figure 19).

Using the present invention, it was shown that KO of St3gal4/6 in combination with Mgat4a/4b/5 interestingly produced biantennary N-glycans without sialylation but with increased poly-LacNAc (Figure 25). Therefore, CHO with st3gal4/6 KO can produce EPO with unprotected LacNAc termini and this opens for the first time the door to de novo engineering of recombinant glycoproteins with α2,6NeuAc protection without competition for endogenous α2,3-sialylation (El Mai, Donadio-Andrei et al. 2013).

Single-antenna N-glycans: Using the present invention, we also investigated the effect of mgat2 KO alone and in combination with Mgat4a/4b/5 and st3gal4/6 (Figure 32). KO of mgat2 alone produced a mixture of single- and biantennary N-glycan structures, and in further combination with Mgat4a/4b/5 homogeneous single-antennary N-glycans. To confirm that the single-antennary N-glycan was linked to the mgat1-controlled p1-2a1-3Man branch, we reintroduced human MGAT4A as well as MGAT4A and 5 in combination with ST6Gal-I by knockin, resulting in tri- and tetraantennary N-glycans, respectively (Figure 32).

Reconstruction of homogeneous glycosylation

The present invention further provides strategies for developing mammalian cell lines with defined and/or more homogeneous glycosylation capabilities that serve as a template for de novo engineering of desirable glycosylation capabilities by introducing one or more glycosyltransferases using site-directed gene insertions and/or classical random transfection of cDNA and/or genomic constructs. The strategy involves inactivating glycosyltransferase genes to obtain homogeneous glycosylation capability in a particular desirable type of glycosylation, for example, using the deconstruction matrix developed herein for N-glycosylation, and for example, but not limited to, inactivating the St3gal4 and 6 sialyltransferase genes in a cell and obtaining a cell without sialic acid protection of the N-glycans. In such a cell, de novo introduction of one or more novel glycosylation capabilities using the more homogeneous truncated glycan product obtained by one or more glycosyltransferase gene inactivation events will provide non-competitive glycosylation and more homogeneous glycosylation by the de novo introduced glycosyltransferases. For example, but not limited to, the introduction of an α2,6-sialyltransferase such as ST6GAL1 into a mammalian cell with inactivated St3gal4 and St3gal6 sialyltransferase genes. The general principle of the strategy is to simplify glycosylation by a particular pathway, e.g., N-glycosylation, to a point where reasonably homogeneous glycan structures are produced in the mammalian cell into which one or more glycosyltransferase gene inactivation events have been introduced in a deconstruction process such as that provided herein for N-glycosylation. Taking such a deconstructed and simplified glycosylation-capable mammalian cell and introducing de novo desirable glycosylation capabilities that are based on the glycan structures produced by the deconstruction.

Reconstruction of N-glycan sialylation by targeted knockin: de novo design of α2,6NeuAc protection

Most circulating glycoproteins in humans are capped with α2,6NeuAc (El Mai, Donadio-Andrei et al. 2013), whereas essentially all recombinant therapeutics are produced with α2,3NeuAc in CHO, HEK293, and other current production cell lines. Therefore, it is clearly desirable to obtain mammalian cell lines such as CHO capable of producing more human-like glycoproteins with a homogeneous α2,6NeuAc coating. Several research groups have addressed this problem and attempted to produce stable CHO cells capable of producing homogeneous 2,6NeuAc-capped N-glycans without success (El Mai, Donadio-Andrei et al. 2013).

To demonstrate the de novo glycoengineering and reconstruction possibilities enabled by our KO deconstruction screen, CHO clones capable of producing homogeneous α2,6NeuAc protection were generated (FIG 25). To avoid past problems with random plasmid integration (El Mai, Donadio-Andrei et al. 2013), a ZFN targeting ST6GAL1 knockin (KI) was used with a modified ObLiGaRe strategy at a SafeHarbor integration site (Maresca, Lin et al. 2013). This strategy also does not involve or require antibiotic selection or maintenance for cloning and stability, which is important for the potential to produce recombinant therapies for clinical use. Human ST6GAL1 introduced into St3gal4/6 knockout CHOs produced the full range of antennal N-glycan structures with a normal degree of NeuAc coating (FIG 25), and when introduced into Mgat4a/4b/5 knockout CHOs, EPO was produced with homogeneous biantennary N-glycans coated by α2,6NeuAc (FIG 25). That the introduced sialylation was indeed α2,6-linked NeuAc and not α2,3-linked was confirmed by negative ion ESI-MS/MS as described for FIGURE 15.

Surprisingly, de novo introduction of ST6Gal-I abolished the lower amounts of poly-LacNAc formed in biantennary structures in Mgat4a/4b/5 knockout CHO (FIG 25). Deletion of B3gnt2 was required to remove poly-LacNAc in both WT and Mgat4a/4b/5 knockout CHO (FIG 18). This example clearly demonstrates the intricate interplay between glycosyltransferases using the same acceptor substrate and how different glycosylation reactions can compete with each other in unpredictable ways. Thus, it is possible to introduce the sialylation capacity of α2,6NeuAc without competing α2,3sialylation, and to reconstruct a CHO cell capable of producing homogeneously coated sialylated N-glycoproteins of α2,6NeuAc with either the normal N-glycan branching repertoire or homogeneous biantennary N-glycans.

Construction of monoantennary N-glycans suitable for homogeneous enzymatic glycomodification

Enzymatic glycoPEGylation and other protein modifications is a well-established method for the site-specific attachment of polyethylene glycol (PEG) chains or other compounds to extend the half-life of drugs (DeFrees, Wang et al. 2006). Site-specific glycomodification of N-glycans using an α2,3-sialyltransferase to introduce the modification via a modified CMP-NeuAc substrate relies on i) a purified recombinant expressed glycoprotein bearing one or more N-glycans; ii) the removal of sialic acids capping the N-glycans on the glycoprotein using a recombinant neuraminidase; and iii) the transfer of the modified sialic acid to exposed galactose residues on the N-glycans using a recombinant sialyltransferase and the modified donor substrate. The N-glycans of recombinant glycoproteins have varying degrees of bi-, tri-, and tetra-antennary structures, and desialylation of recombinant glycoproteins exposes multiple galactose residues from each antennal structure, resulting in substantial heterogeneity in the enzymatic glycomodification process. To overcome the desialylation step, we produced a series of CHO-engineered designs that produce uncapped N-glycans (Figs. 19, 25, and 32), which would allow direct enzymatic glycomodification without neuraminidase pretreatment to remove sialic acids. To overcome the heterogeneity induced by the multiple galactose acceptor sites exposed for enzymatic transfer on each multi-antenna N-glycan, we targeted mgat2 to block the formation of biantennary structures, but knockout of mgat2 alone did not generate homogeneous monoantennary structures as originally expected (FIG 32). Instead, approximately 50% remained biantennary, but combined with knockout of mgat4A/4B/5, fairly homogeneous monoantennary structures were obtained. The only detectable heterogeneity was associated with poly-LacNAc, which does not provide exposure of multiple galactose residues (FIG 18).

A fairly homogeneous monoantennary structure can also be achieved by the following double KOs: Mgat2/4B KO or Mgat2/5 KO. (Figure 35).

For the IgG antibody, a homogeneous mono-antennal structure and a higher degree of sialylation can be obtained by KO of Mgat2 (Figure 36).

Elimination of unnecessary glycosylation pathways

The methods of the invention allowed to investigate whether inactivation of different O-glycosylation pathways would affect cell viability and glycosylation efficiencies of non-targeted glycosylation pathways. First, we abrogated the Core1 elongation pathway of O-GalNAc glycosylation as previously described (Yang, Halim et al. 2014) by inactivating the Cosmc chaperone for core1 synthase, C1Gal-T1, leading to CHO cells capable of producing EPO with truncated O-GalNAc glycosylation limited to a single GalNAc residue without sialylation at the single O-glycosite. WT CHO produces α2,3- and α2,6-sialic acid-capped O-glycans, and O-GalNAc glycosylation is one of the most abundant types of protein glycosylation in a cell and therefore demanding on the sugar nucleotide pool. Consistent with this, we found that EPO produced in CHO with Cosmc knockdown and O-glycan elongation produced EPO with enhanced sialic acid coating of N-glycans at all three N-glycosites present in EPO (FIG 29).

Using the methods of the present invention, the O-Man glycosylation elongation pathway was targeted by knockdown of Pomgnt1, which encodes the p2GlcNAc transferase that synthesizes the major pathway of O-Man glycan elongation (PNAS). Dystroglycan-like proteins expressed in mammalian cells, including CHO (Yang, Halim et al. 2014), have a dense coating of the O-Man tetrasaccharide (NeuAca2,3Galp1,4GlcNAcp1,2Mana). When EPO was expressed in a Pomgnt1 knockdown CHO cell, the degree of protection was enhanced by galactose, polyLacNac, and sialic acid (FIG 30).

The present invention was applied to inactivate the O-Xyl glycosylation elongation pathway by KO of B4galt7, which encodes the p4Galactosyltransferase that synthesizes the second step in the linker region required for proteoglycan biosynthesis (Almeida, Levery et al. 1999). CHO cells produce proteoglycans and have efficient capabilities to produce them from, for example, exogenously supplied saccharides. When EPO was expressed in a B4galt7 knockout CHO cell, the degree of coating of galactose, poly-LacNac and sialic acid was increased (FIG. 30).

The present invention was also used to inactivate combinations of two or three genes that control the O-GalNAc, O-Man, and O-Xyl glycosylation pathways. Specifically, CHO cells with knockout of B4galt7 and Pomgnt1, and with knockout of B4galt7, Pomgnt1, and Cosmc were generated. When EPO was expressed in such CHO cell lines with knockout of two or three types of O-glycosylation pathways, the degree of galactose and sialic acid protection was enhanced.

These results clearly demonstrate that the more homogeneous glycosylation capacity of recombinantly expressed proteins in mammalian cells such as CHO can be improved by inactivating unnecessary glycosylation pathways that use up common pools of sugar nucleotide donors and other parts of the cells' general metabolic and glycosylation capabilities. Those skilled in the art are aware that inactivation of other genes listed in TABLES 1 and 2 in the above glycosylation pathways or other glycosylation pathways will have similar effects and may be desirable engineering events alone or in combination with any of the genetic engineering performed herein or conceivable from the glycosylation pathways described in FIG. 13.

The general principle for selecting optimal inactivation strategies to direct sugar nucleotide usage in a mammalian host cell to glycosylation pathways relevant to the recombinant production of a given glycoprotein is to block non-relevant glycosylation pathways at an early step in the biosynthetic pathway where a single gene inactivation results in truncated glycan structures preferably without elongation and capping by sialic acid, and/or galactose, and/or GlcNAc. Examples are provided for the inactivation of three distinct types of O-glycosylation by targeting the second step in their biosynthesis to avoid and/or limit the use of CMP-NeuAc, UDP-Gal and UDP-GlcNAc for the elongation of these pathways. It is evident that many other genes listed in Table 2 with assigned biosynthetic steps for the different glycosylation pathways in the mammalian CHO cell, can be targeted for inactivation to achieve the same or similar effects using donor sugar nucleotides taking into account the effects that inactivation has on the glycosylation pathway and altered glycan structures.

Elimination of unnecessary glycosyltransferases

Resident ER-Golgi glycosyltransferases are maintained for secretory function by a variety of partially understood functions including kin recognition, transmembrane spanning type and length, interactions across the stalk (Colley 1997), and through COP-I vesicle retrograde transport presumably dependent on the cytosolic tails and TMs of glycosyltransferase proteins, e.g., GOLPH3 (Eckert, Reckmann et al. 2014). When glycosyltransferases are not retained in their natural position in the secretory pathway, they may function differently and adversely in glycosylation, and when glycosyltransferases lose their hold, e.g., by proteolytic cleavage at the stalk, they are secreted and released with little or no influence on glycosylation processes in the cells. Considering the numerous glycosyltransferases expressed in any mammalian cell, and for example at least more than 60 in CHO cells (TABLE 3), we hypothesized that there would be common retention mechanisms for groups and/or classes of enzymes. A group or class of enzymes could be constituted without any intention of limitations, for example, enzymes that work in the same pathway and/or in consecutive biosynthetic steps, and/or enzymes retained in the same subcellular topology, and/or enzymes that have a similar amino acid sequence and/or structural retention signals. The latter could, without any intention of limitations, be found, for example, in homologous isozymes since they have similar sequences, related catalytic properties and, at least in some cases, have been shown to have similar subcellular topologies (Rottger, White et al. 1998). One group of such isoenzymes are the β-galactosyltransferases, of which B4GAL-T1, T2, T3, and T4 have the highest degree of sequence similarity and related functions (Amado, Almeida et al. 1999; Togayachi, Sato et al. 2006). The understanding of the in vivo functions of these four enzymes is poor, and all have been shown to produce the Galp1-4GlcNAc linkage in vitro. As shown here, p4Galt1 and 3 appear to be the major β-galactosyltransferases involved in N-glycosylation, and previous studies have shown that B4GALT4 has functions mainly in glycolipid biosynthesis and/or in sulfated glycans (Schwientek, Almeida et al. 1998). Thus, we generated B4galt4 knockout CHO cells in WT CHO as well as in combination with mgat4a/4b/5 knockout. EPO expression in B4galt4 knockout CHO cells alone resulted in increased tetraantennary N-glycans with a more complete sialic acid coating (Fig. 17). B3gnt2 knockout resulted in EPO glycosylates lacking poly LacNac (Fig. 35) whereas B4galt5/6 knockout gave more poly LacNac (Fig. 35). In this line, IgG produced in St3gal4/6 deleted CHO cells displayed higher galactosylation (Fig. 36). These results clearly demonstrate that the capacity for more homogeneous glycosylation of recombinantly expressed proteins in mammalian cells such as CHO can be enhanced by inactivating unnecessary glycosyltransferases that may compete for mechanisms directing ER-Golgi residency and/or use up common pools of sugar nucleotide donors and other parts of the overall metabolic and glycosylation capabilities in a mammalian cell.

These results clearly demonstrate that the more homogeneous glycosylation capacity of recombinantly expressed proteins in mammalian cells such as CHO can be improved by inactivating unnecessary glycosylation pathways that use up common pools of sugar nucleotide donors and other parts of the cells' general metabolic and glycosylation capabilities. Those skilled in the art are aware that inactivation of other genes listed in TABLES 1 and 2 in the above glycosylation pathways or other glycosylation pathways will have similar effects and may be desirable engineering events alone or in combination with any of the genetic engineering performed herein or conceivable from the glycosylation pathways outlined in FIG. 13.

Elimination of post-translational N-glycosylation capacity

The oligosaccharyltransferase complex that initiates protein N-glycosylation consists of eight proteins, one subunit of which involves two paralogous genes, STT3A and STT3B. Studies have suggested that an oligosaccharyltransferase complex containing the STT3A subunit functions primarily in cotranslational N-glycosylation, whereas an oligosaccharyltransferase complex containing the STT3B subunit functions primarily in posttranslational N-glycosylation. We hypothesize that the fidelity and stoichiometry of cotranslational N-glycosylation are naturally better than those of posttranslational N-glycosylation. Many recombinantly expressed glycoproteins have misused N-glycans, resulting in incomplete N-glycan occupancy (stoichiometry) and/or N-glycan processing (structure). Therefore, we next generated Stt3b knockout CHO cells as well as combined Mgat4A/4B/5 knockout CHO cells. These CHO cells have an enhanced ability to produce homogeneous glycoproteins with a higher degree of N-glycan stoichiometry and/or glycan structures when expressing proteins with naturally underused N-glycosites.

Selective elimination strategy

It is clear to those skilled in the art that inactivation of a glycosyltransferase gene can have a multitude of outcomes and effects on the transcription and/or the protein product translated from it. The targeted inactivation experiments performed here involved PCR and sequencing of the alterations introduced into the genes, as well as RNA sequencing analysis of clones to determine whether a transcript was formed and whether potential novel splicing variations introduced new protein structures. In addition, there are methods for determining the presence of proteins from such transcripts, including mass spectrometry and SDS-PAGE Western blot analysis with relevant antibodies that detect the most N-terminal region of the protein products.

Selection constructs were generally designed to target the first 1/3 of the open reading frame (ORF) of the coding regions, but other regions were also targeted (Figure 14B). For most of the targeted clones, out-of-frame mutations (indels) that introduced stop or nonsense codons within the same exon were selected. This would be expected to produce truncated proteins, if any, lacking the catalytic domain and therefore no enzymatic activity. Most KO clones had insertions and/or deletions (indels) in the range of ±20 bps, and most of the target genes were present with two alleles, while some (mgat4B and mgat5) were present with one or three alleles, respectively (TABLE 4). In some cases, larger deletions were found that altered one or more exons and exon/intron boundaries, also resulting in truncated proteins.

Most ER-Golgi glycosyltransferases share the common type 2 transmembrane structure with a short cytosolic tail that can direct retrograde trafficking and residence time, an uncleaved signal peptide containing a hydrophobic transmembrane α-helix domain for retention at the ER-Golgi membrane, a variable length stem region thought to shift the catalytic domain into the lumen of the ER-Golgi, and a C-terminal catalytic domain required for enzyme function (Colley 1997). Only polypeptide GalNAc transferases have an additional C-terminal lectin domain (Bennett, Mandel et al. 2012). The genomic organization of glycosyltransferase genes varies substantially: some genes have a single coding exon and others have more than 10-15 coding exons, although a few glycosyltransferase genes produce different splice variants encoding different protein products.

Inactivation of glycosyltransferase genes in early coding regions can have a multitude of effects on the transcription and protein products if these are carried out: i) one or more transcripts may be unstable and rapidly degraded resulting in little or no transcription and/or protein; ii) one or more transcripts may be stable but untranslated or only mistranslated resulting in little or no protein translation; iii) one or more transcripts may be stable and translated resulting in protein translation; iv) one or more transcripts may result in protein translation but the protein products are degraded due to, for example, truncations and/or misfolding; v) one or more transcripts may result in protein translation and stable protein products that are truncated and enzymatically inactive; and vi) one or more transcripts may result in protein translation and stable protein products that are truncated but have enzymatic activity.

It is apparent to those skilled in the art that gene inactivation leading to protein products with enzymatic activity is undesirable, and these events are readily detected by the methods used in the present invention, for example, but not limited to, lectin/antibody labeling and glycolytic profiling of proteins expressed in mutant cell lines.

However, it is desirable to eliminate potential truncated protein products that may be expressed from mutated transcripts, as these may have undesirable effects on glycosylation capacity. Therefore, truncated protein products of type 2 glycosyltransferase genes containing part or all of the cytosolic domain and/or the transmembrane retention signal and/or the stem region and/or part of the catalytic domain may exert several effects on glycosylation in cells. For example, the cytosolic tail may compete for COP-I retrograde trafficking of Golgi-resident proteins (Eckert, Reckmann et al. 2014), the transmembrane domain may compete for localization to the ER-Golgi and possible normal protein associations and/or aggregations, and the stalk region, as well as part of an inactive catalytic domain, may have similar or part of functions in normal protein associations and/or aggregations in the ER-Golgi. These and other unknown functions may affect specific glycosylation pathways in which the enzymes participate, specific functions of the isozymes, or, more generally, the glycosylation capabilities of a cell. While these functions and effects are currently unknown and unpredictable, it is an inherent part of the present invention that the selection of mammalian cell clones with inactivated glycosyltransferase genes includes the selection of editing events that do not produce truncated protein products.

It is evident that the performed KO deconstruction screen identifies key glycogens that control decisive steps in protein N-glycosylation in CHO and provides a design matrix for the genetic deconstruction of N-glycosylation capacity to desirable homogeneous structures that can serve as starting points for the reconstruction of improved, novel, and/or more homogeneous glycosylation capabilities by the stable introduction of one or more glycosyltransferase genes. FIGURE 28 provides general examples of N-glycan scaffolds that can be produced by the deconstruction design matrix, which serve as platforms for the reconstruction of desirable more homogeneous glycosylation capabilities. The value of this combined deconstruction and reconstruction strategy is clearly exemplified here by the design of EPO-producing CHO cells with homogeneous biantennary N-glycans with α2,3 and, for the first time, homogeneous N-glycans capped with α2,6-sialic acid. These CHO lines clearly complement and extend previous efforts to establish glycoengineered yeast capable of EPO production with homogeneous biantennary N-glycans, as well as the highly successful chemical synthesis of EPO with biantennary N-glycans (Hamilton, Davidson et al. 2006; Wang, Dong et al. 2013). More specifically, the CHO lines produced enable the production of EPO and other glycoproteins in a proven mammalian host cell with well-established protein folding and processing capabilities and a decades-long safety profile for human therapies.

It is further evident from the KO deconstruction screen performed here that the CHO cell has remarkable plasticity and tolerance to glycoengineering. The glycosylation capabilities achieved by the KO lines in CHO were found to be stable and consistent in the production of recombinant glycoproteins, and these engineered CHO lines are expected to perform similarly to WT CHO lines in bioprocessing. Furthermore, the reconstruction of glycosylation capabilities by KI exemplified by the introduction of ST6Gal-1 without competing α2,3-sialyltransferases demonstrates that stable and consistent glycoengineering can be achieved by the deconstruction and reconstruction strategy invented here, and this should be applicable to other desirable glycosylation capabilities. This includes reconstructing the endogenous glycosylation capabilities of WT CHO, which are inconsistent and heterogeneous, such as the formation of multi-antennary N-glycan branches (tri- and tetra-antennary) and poly-LacNAc chains. Other examples include improving the ability to protect N-glycans from IgG. Furthermore, the engineered design matrix is expected to be transferable to any CHO host line, as well as to established production lines. Furthermore, the engineered design matrix is expected to be transferable to any mammalian cell line with appropriate expansion depending on the existence of more complex glycosylation capabilities. The panel of CHO cells generated here will allow for de novo dissection and reconstruction design of N-glycosylation for any protein therapeutic. Furthermore, this panel of CHO cells can be used to produce therapeutic glycoproteins with a range of different glycoconformities, which will for the first time allow the biological properties of defined glycoconformities to be experimentally assessed and optimal designs to be rationally selected. The CHO cell panel produced here allows direct testing of the functional effects of specific N-glycan structures on any therapeutic glycoprotein, which will provide guidance for the optimal design of a glycoprotein drug. Thus, using the CHO cell panel, the effect of N-glycan branching, poly-LacNAc extension, and sialylation type can be systematically addressed. These effects include, for example, protein secretion, stability, yield, blood circulatory half-life, biodistribution, bioactivity, and general pharmacokinetic properties relevant to therapeutic drugs. The CHO panel will also allow, for example, the design of de novo glycoprotein therapies, where the introduction of N-glycans is used to improve circulatory half-life without interfering with biological activity. In this case, the effect of the position, size, and structure of the introduced N-glycans can be mapped in detail to design new glycoprotein biologics with optimal pharmacokinetic properties. The present invention thus provides a new era for the largest producer of glycoprotein therapies: the CHO cell.

In a further embodiment of the present invention, mgat4A, mgat4B, mgat5, and FUT8 have been deleted. In one embodiment of the present invention, the cell or cell line is a mammalian cell.

In another embodiment of the present invention, the cell is derived from Chinese hamster ovary or human kidney. In yet another embodiment of the present invention, the cell is selected from the group consisting of CHO, NS0, SP2/0, YB2/0, CHO-K1, CHO-DXB11, CHO-DG44, CHO-S, HEK293, HUVEC, HKB, PER-C6, NS0, or derivatives of any of these cells.

In one embodiment of the present invention the cell is a CHO cell.

In another embodiment of the present invention, the cell further encodes an exogenous protein of interest. In another embodiment of the present invention, the protein of interest is an antibody, an antibody fragment, or a polypeptide.

In another embodiment of the present invention, the antibody is an IgG antibody.

In one embodiment of the present invention is the EPO polypeptide, a protein involved in hemostasis, including a coagulation factor.

In one embodiment of the present invention glycosylation is made more homogeneous.

In another embodiment of the present invention the glycosylation is non-sialylated.

In another embodiment of the present invention, the glycosylation is not galactosylated.

In one embodiment of the present invention it comprises biantennary N-glycosylation.

In another embodiment of the present invention, the glycosylation does not comprise poly-LacNAc.

In another embodiment of the present invention, glycosylation is any combination without fucose.

In one embodiment of the present invention, B4galt1 has been eliminated allowing the generation of more homogeneous N-glycans without galactose.

In another embodiment of the present invention, B4galt1 and fut8 have been eliminated allowing the generation of more homogeneous N-glycans without galactose and fucose.

In yet another embodiment of the present invention, B4galt1 and B4galt3 have been deleted, resulting in more homogeneous galactose-free N-glycans than those obtained with a single deletion of B4galt1 or B4galt3 (Figure 36).

In one embodiment of the present invention, the cell has knock out of one or more of the glycosyl transferase genes selected from the group consisting of mgat1, mgat2, mgat4A, mgat4B, mgat4C, mgat5 and mgat5B.

In another embodiment of the present invention, the cell has knockout of one or more of the glycosyltransferase genes selected from the group consisting of mgat4A, mgat4B, and mgat4C.

In a further embodiment of the present invention, the cell has been knocked out of mgat4A, mgat4B, and mgat4C. In another embodiment of the present invention, the cell comprises the deletion of mgat5 and/or mgat5B.

In another embodiment of the present invention, the cell comprises knockout of mgat5 and mgat5B.

In one embodiment of the present invention, the cell comprises the deletion of one or more of the galatosylation genes selected from the group consisting of B4gal1, B4gal2, B4gal3 and B4gal.

In another embodiment of the present invention, B4gal1 or B4gal3, or B4gal1 and B4gal3, have also been deleted. In a further embodiment of the present invention, the cell also has knockout of one or more of the poly-LacNAc elongation genes selected from the group consisting of B3gnt1, B3gnt2, and B3gnt8. In yet another embodiment of the present invention, B3gnt2 has also been deleted.

In one embodiment of the present invention, B3gnt2, mgat4A, mgat4B, and mgat5 have also been deleted. In another embodiment of the present invention, the cell also has knockout of one or more of the sialyltransferase genes selected from the group consisting of ST3gal3, ST3gal4, and ST3gal6.

In yet another embodiment of the present invention, ST3gal3, ST3gal4, and ST3gal6 have also been deleted. In a further embodiment of the present invention, ST3gal4 and ST3gal6 have also been deleted.

In one embodiment of the present invention, ST3gal4, ST3gal6, mgat4A, mgat4B and mgat5 have also been deleted.

In another embodiment of the present invention, ST6gal1 has also been deleted and ST3gal4 and ST3gal6 have been deleted.

In yet another embodiment of the present invention, ST6gal1 has been introduced and ST3gal4, ST3gal6, mgat4A, mgat4B and mgat5 have been deleted.

In one embodiment of the present invention, mgat2, ST3gal4 and ST3gal6 have been deleted.

In another embodiment of the present invention, mgat2, mgat4A, mgat4B and mgat5 have been deleted.

In another embodiment of the present invention, mgat2, ST3gal4, ST3gal6, mgat4A, mgat4B and mgat5 have been deleted.

An aspect of the present invention relates to a glycoprotein according to the present invention, which is a homogeneous glycoconjugate produced from a glycoprotein having a simplified glycan profile. An object of the present invention is to provide a cell capable of expressing a gene encoding a polypeptide of interest, wherein the polypeptide of interest is expressed comprising one or more post-translational modification patterns.

In one embodiment of the present invention, the post-translational modification pattern is glycosylation. The optimal glycoform of a glycoprotein can be identified by the following process: (i) producing a plurality of different glycoforms of said glycoprotein by expression in cell lines harboring at least one novel glycosylation capability, for example, harboring two or more modifications of the expression levels of the GT gene, and (ii) determining the activity of the different glycoforms compared to a reference glycoprotein in (a) suitable bioassay(s); and iii) selecting the glycoform with the highest or greatest activity and determining the genotype fingerprint of the production cell that correlates with the highest or greatest activity level of said glycoprotein.

The process described above allows the identification of the optimal glycoform of a glycoprotein. For those skilled in the art, by using the genotypic fingerprint identified in (iii), an efficient engineered cell line with the optimal genotype for producing glycoprotein with said glycoform can be generated. Another aspect of the present invention relates to a glycoprotein according to the present invention, which is a homogeneous glycoprotein conjugate comprising a polymer selected from PEG, HEP, XTEN, PSA, HES.

A further aspect of the present invention relates to a glycoprotein according to the present invention, which is a homogeneous PEGylated protein conjugate produced from a protein variant according to the invention having a simplified glycan profile.

Another aspect of the present invention relates to a glycoprotein according to the present invention, which is a homogeneous PEGylated EPO conjugate produced from an EPO variant having a simplified glycan profile.

In one embodiment of the present invention, Cosmc is also introduced or removed.

In one embodiment of the present invention, the resulting glycosylation has one or more of the following changes selected from the group consisting of a simpler glycan structure, a more homogeneous product, more sialic acids per molecule, non-sialylated, non-galactosylated, more homogeneous bi-antennary, more homogeneous mono-antennary, more homogeneous tri-antennary, more homogeneous without poly-LacNAc, increased productivity in cell culture, new stable homogeneous glycosylation capabilities, more human glycosylation, more homogeneous glycosylation and improved IgG ADCC targeting, modified fucose level, no fucose, improved substrate for generating glycoconjugates.

In another aspect of the present invention, one or more of the aforementioned genes are deleted using transcription activator-like effector nucleases (TALENs).

TALENs are artificial restriction enzymes generated by fusing a TAL effector DNA binding domain to a DNA cleavage domain.

In yet another aspect of the present invention, one or more of the aforementioned genes are deleted using CRISPR (clustered regularly interspaced short palindromic repeats).

CRISPRs are DNA loci containing short repeats of base sequences. Each repeat is followed by short segments of “spacer DNA” from previous exposures to a virus.

CRISPRs are often associated with cas genes, which encode CRISPR-related proteins. The CRISPR/Cas system is a prokaryotic immune system that confers resistance to foreign genetic elements such as plasmids and phages and provides a form of acquired immunity.

CRISPR spacers recognize and cut these foreign genetic elements in a manner analogous to siRNA in eukaryotic organisms.

The CRISPR/Cas system is used for gene editing (adding, altering, or changing the sequence of specific genes) and gene regulation in species across the tree of life. By introducing the Cas9 protein and the appropriate guide RNAs into a cell, the organism's genome can be cut at any desired location.

In a further embodiment of the present invention, the mammalian cell has or does not have α-1,6-fucosyltransferase activity.

The cell of the present invention may be a cell that does not contain the gene of interest to be expressed, or the cell may contain the gene of interest to be expressed. A cell that does not contain the gene of interest to be expressed is often referred to as a "naked cell."

The host cell of the present invention can be any host, provided it can express an antibody molecule. Examples include a yeast cell, an animal cell, an insect cell, a plant cell, and the like.

In one embodiment of the present invention, the cell is selected from the group consisting of CHO, NS0, SP2/0, YB2/0, YB2/3HL.P2.G11.16Ag.20, n So , SP2/0-Ag14, BHK cell derived from a Syrian hamster kidney tissue, antibody-producing hybridoma cell, human leukemia cell line (Namalwa cell), an embryonic stem cell, and a fertilized egg.

In a preferred embodiment of the present invention, the cell is a CHO cell.

The cell may be an isolated cell or in cell culture or a cell line.

The protein of interest can be various types of protein, and in particular proteins that benefit from being expressed as glycoproteins, and which can be produced by the cells of the present invention. In a preferred embodiment, the protein is erythropoietin (EPO).

In another embodiment is the protein a1-antitrypsin.

In one aspect of the present invention, the protein is a recombinant blood factor.

In one embodiment of the present invention, the recombinant blood factor is selected from the group consisting of one or more of factors VIII, IX, factor XIII subunit A, thrombin, and VIIa.

In another embodiment of the present invention, the protein of interest is a coagulation factor, such as coagulation factor II (FII), coagulation factor V (FV), coagulation factor VII (FVIIa), coagulation factor VIII (FVIII), coagulation factor IX (FIX), coagulation factor X (FX), or coagulation factor XIII (FXIII).

In one aspect of the present invention is the protein human growth hormone.

In another embodiment of the present invention, the protein of interest is an antibody, an antibody fragment, such as a Fab fragment, an Fc domain of an antibody, or a polypeptide.

A further aspect of the present invention relates to a method for producing an enzymatically modified glycoprotein, comprising the step of generating a cell with specific glycosylation properties, and the enzymatic modification of interest.

In one embodiment of the present invention, enzymatic modification of a polymer is performed.

In another embodiment of the present invention, the enzymatic modification is a conjugation to another protein. Another aspect of the present invention relates to a glycoprotein according to the present invention, which is a homogeneous glycoprotein conjugate comprising a polymer.

The glycoprotein can be enzymatically modified with the polymer, and the glycoprotein can be produced by the cell according to the present invention.

In one aspect of the present invention, the protein is a recombinant thrombolytic, anticoagulant, or other blood-related product.

In one embodiment of the present invention is the recombinant thrombolytic, anticoagulant or other blood-related product selected from the group consisting of one or more of tissue plasminogen activator (tPA), hirudin, antithrombin, plasmin, plasma kallikrein inhibitor and activated protein C.

In one aspect of the present invention the protein is a recombinant hormone.

In one embodiment of the present invention, the recombinant hormone is selected from the group consisting of one or more of the following: insulin, insulin degludec, human growth hormone, somatropin, pegvisomant, follicle-stimulating hormone, follitropin alfa, corifollitropin alfa, follitropin beta, metreleptin, liraglutide, parathyroid hormone, lutropin, teriparatide, nesiritide, and glucagon.

In one aspect of the present invention, the protein is a recombinant growth hormone.

In one embodiment of the present invention, the recombinant growth hormone is selected from the group consisting of one or more of EPO, filgrastim, sargramostim, mecaserim, and palifermin.

In one aspect of the present invention, the protein is a recombinant interferon, interleukin or tumor necrosis factor.

In one embodiment of the present invention is the recombinant interferon, interleukin or tumor necrosis factor selected from the group consisting of one or more of interferon alpha, PEGinterferon alpha, PEGinterferon alpha-2a, PEGinterferon alpha, interferon beta-1b, algulcosidase alpha and laronidase.

In one aspect of the present invention, the protein is an antigen or vaccine component.

In one aspect of the present invention, the protein of the present invention is relevant to a disease or disorder selected from the group consisting of one or more of the following: hemophilia A, hemophilia B, acute myocardial infarction, heparin-associated thrombocytopenia, venous thrombosis, symptomatic vitreomacular adhesion/vitreomacular traction, acute angioedema, hereditary antithrombin deficiency, hereditary angioedema, sepsis, diabetes, diabetes mellitus, growth retardation/growth hormone deficiency, infertility/subfertility, type 2 diabetes, osteoporosis, hypoglycemia, Paget's disease, cancer, anemia, neutropenia, hepatitis C and hyperuricemia, rheumatoid arthritis, hepatitis A and B, arthritis, colitis, Crohn's disease, psoriasis, ankylosing spondylitis, ulcerative colitis.

In one embodiment of the present invention, the protein is an antibody.

In a preferred embodiment of the present invention, the antibody is an IgG antibody.

In the present invention, the antibody molecule includes any molecule, as long as it comprises the Fc region of an antibody. Examples include an antibody, an antibody fragment, a fusion protein comprising an Fc region, and the like.

An antibody is a protein produced in the body by an immune reaction as a result of stimulation by an exogenous antigen and has the ability to specifically bind to the antigen. Examples of antibodies include antibodies secreted by hybridoma cells prepared from spleen cells of animals immunized with antigens; antibodies prepared by genetic recombination techniques, i.e., antibodies obtained by introducing an antibody expression vector containing an antibody gene into a host cell; and the like. Specific examples include antibodies produced by hybridoma cells, humanized antibodies, human antibodies, and the like.

A hybridoma is a cell obtained by cell fusion between a B cell obtained by immunizing a mammal other than a human with an antigen and a mouse-derived myeloma cell or similar and can produce a monoclonal antibody having the desired antigen specificity.

Examples of the humanized antibody include a human chimeric antibody, a CDR-grafted human antibody, and the like.

A human chimeric antibody is an antibody comprising an antibody heavy chain variable region (hereinafter referred to as "HV" or "VH", the heavy chain being "H chain") and an antibody light chain variable region (hereinafter referred to as "LV" or "VL", the light chain being "L chain"), both of an animal other than human, a human antibody heavy chain constant region (hereinafter also referred to as "CH") and a human antibody light chain constant region (hereinafter also referred to as "CL"). As the animal other than human, any animal such as mouse, rat, hamster, rabbit or the like can be used, as long as a hybridoma can be prepared therefrom.

The human chimeric antibody can be produced by obtaining cDNA encoding VH and VL from a monoclonal antibody-producing hybridoma, inserting them into an expression vector for a host cell having genes encoding the human antibody CH and the human antibody CL to thereby construct a human chimeric antibody expression vector, and then introducing the vector into a host cell to express the antibody.

As the CH of the human chimeric antibody, any CH can be used, as long as it belongs to the human immunoglobulin (hereinafter referred to as "hIg"). However, those belonging to the hIgG class are preferable, and any of the subclasses belonging to the hIgG class, such as hIgG1, hIgG2, hIgG3, and hIgG4, can be used. In addition, as the CL of the human chimeric antibody, any CL can be used, as long as it belongs to the hIg class, and those belonging to the k class or the A class can also be used.

A human CDR-grafted antibody is an antibody in which the VH and VL CDR amino acid sequences of an antibody derived from an animal other than a human are grafted into appropriate positions of the VH and VL of a human antibody.

The CDR-grafted human antibody can be produced by constructing cDNAs encoding V regions in which the VH and VL CDRs of an antibody derived from an animal other than human are grafted into VH and VL CDRs of a human antibody, inserting them into an expression vector for a host cell having genes encoding the human antibody CH and the human antibody CL to thereby construct a CDR-grafted human antibody expression vector, and then introducing the expression vector into a host cell to express the CDR-grafted human antibody.

As the CH of the CDR-grafted human antibody, any CH may be used as long as it belongs to the hIg class, but those of the hIgG class are preferable, and any of the subclasses belonging to the hIgG class, such as hIgG1, hIgG2, hIgG3, and hIgG4, may be used. In addition, as the CL of the CDR-grafted human antibody, any CL may be used as long as it belongs to the hIg class, and those belonging to the k class or the A class may also be used.

A human antibody is originally an antibody that naturally exists in the human body, but also includes antibodies obtained from a human antibody phage library, a human antibody-producing transgenic animal, and a human antibody-producing transgenic plant, which are prepared based on recent progress in genetic engineering, cell engineering, and developmental engineering.

With respect to the antibody existing in the human body, a lymphocyte capable of producing the antibody can be cultured by isolating a human peripheral blood lymphocyte, immortalizing it by infecting it with the EB virus or the like and then cloning it, and the antibody can be purified from the culture.

The human antibody phage library is a library in which antibody fragments such as Fab, single chain antibodies and the like are displayed on the surface of the phage by inserting a gene encoding an antibody prepared from a human B cell into a phage gene. A phage displaying an antibody fragment with the desired antigen-binding activity can be recovered from the library, using its binding activity to an antigen-immobilized substrate as a marker. The antibody fragment can be further converted into a human antibody molecule comprising two complete B chains and two complete L chains by genetic engineering techniques.

A human antibody-producing transgenic non-human animal is an animal in which a human antibody gene is introduced into the cells. Specifically, a human antibody-producing transgenic animal can be prepared by introducing a human antibody gene into an ES cell of a mouse, transplanting the ES cell into an early-stage embryo of another mouse, and then developing it. The transgenic animal can also be prepared by introducing a human chimeric antibody gene into a fertilized egg and developing it. As for the method of preparing a human antibody from the human antibody-producing transgenic animal, the human antibody can be produced and accumulated in culture to obtain a human antibody-producing hybridoma by a hybridoma preparation method commonly performed in mammals other than humans and then culturing it.

Examples of transgenic non-human animals include cattle, sheep, goats, pigs, horses, mice, rats, birds, monkeys, rabbits and the like.

Another aspect of the present invention relates to a method for producing an antibody composition, comprising culturing the mammalian cell according to the present invention in a medium for producing and accumulating an antibody composition in the culture; and recovering the antibody composition from the culture.

Furthermore, in the present invention, it is preferable that the antibody is an antibody that recognizes an antigen related to a tumor, an antibody that recognizes an antigen related to an allergy or inflammation, an antibody that recognizes an antigen related to a circulatory organ disease, an antibody that recognizes an antigen related to an autoimmune disease, or an antibody that recognizes an antigen related to a viral or bacterial infection, and a human antibody belonging to the IgG class is preferable.

An antibody fragment is a fragment comprising the Fc region of an antibody. Examples of the antibody fragment include an H chain monomer, an H chain dimer, and the like.

A fusion protein comprising an Fc region is a composition in which an antibody comprising the Fc region of an antibody or antibody fragment is fused with a protein such as an enzyme, a cytokine or the like.

An embodiment of the present invention relates to a composition comprising the antibody of the present invention, also referred to as an antibody composition.

In one embodiment of the present invention, the antibody or antibody composition exhibits high ADCC activity.

In the present invention, the ADCC activity is a cytotoxic activity in which an antibody bound to a cell surface antigen on a tumor cell in the living body activates an effector cell through an Fc receptor existing on the Fc region of the antibody and the surface of the effector cell and thereby obstructs the tumor cell and the like.

The antibody composition of the present invention has potent antibody-dependent cell-mediated cytotoxic (ADCC) activity. An antibody having potent antibody-dependent cell-mediated cytotoxic activity is useful for preventing and treating various diseases, such as cancers, inflammatory diseases, immune diseases such as autoimmune diseases, allergies, and the like, circulatory organ diseases, and viral or bacterial infections.

In the case of cancers, that is, malignant tumors, cancer cells grow. General antitumor agents inhibit cancer cell growth. In contrast, an antibody with potent antibody-dependent cell-mediated cytotoxic activity can treat cancers by damaging cancer cells through its cell-killing effect and is therefore more effective as a therapeutic agent than general antitumor agents.

In immune diseases such as inflammatory diseases, autoimmune diseases, allergies, and the like, the in vivo reactions of the diseases are induced by the release of a mediator molecule by immunocytes, so that the allergic reaction can be inhibited by eliminating immunocytes by an antibody having potent antibody-dependent cell-mediated cytotoxic activity.

Examples of circulatory organ diseases include arteriosclerosis and the like. Currently, arteriosclerosis is treated with balloon catheters, but circulatory organ diseases can be prevented and treated by inhibiting arterial cell growth in the restricted area after treatment using an antibody with potent antibody-dependent cell-mediated cytotoxic activity.

Various diseases, including viral and bacterial infections, can be prevented and treated by inhibiting the proliferation of cells infected with a virus or bacteria using an antibody that has potent antibody-dependent cell-mediated cytotoxic activity.

The glycoproteins of the present invention may therefore be used to treat immune diseases, cancer, viral or bacterial infections or other diseases or disorders mentioned above.

The medicament comprising the glycoprotein composition of the present invention may be administered as a therapeutic agent alone, but generally, it is preferable to provide it as a pharmaceutical formulation produced by an appropriate method well known in the technical field of manufacturing pharmacy, by mixing it with at least one pharmaceutically acceptable carrier.

It is desirable to select a route of administration that is most effective for treatment. Examples include oral administration and parenteral administration, such as buccal, tracheal, rectal, subcutaneous, intramuscular, intravenous, or similar. In an antibody preparation, intravenous administration is preferable.

The dosage form includes sprays, capsules, tablets, granules, syrups, emulsions, suppositories, injections, ointments, tapes and the like.

Examples of the pharmaceutical preparation suitable for oral administration include emulsions, syrups, capsules, tablets, powders, granules and the like.

Liquid preparations, such as emulsions and syrups, can be produced using, as additives, water; saccharides, such as sucrose, sorbitol, fructose, etc.; glycols, such as polyethylene glycol, propylene glycol, etc.; oils, such as sesame oil, olive oil, soybean oil, etc.; antiseptics, such as p-hydroxybenzoic acid esters, etc.; flavors, such as strawberry flavor, mint flavor, etc.; and the like.

Capsules, tablets, powders, granules, and the like can be produced by using, as additives, fillers such as lactose, glucose, sucrose, mannitol, etc.; disintegrating agents such as starch, sodium alginate, etc.; lubricants such as magnesium stearate, talc, etc.; binders such as polyvinyl alcohol, hydroxypropyl cellulose, gelatin, etc.; surfactants such as fatty acid esters, etc.; plasticizers such as glycerin, etc.; and the like.

Examples of pharmaceutical preparation suitable for parenteral administration include injections, suppositories, sprays and the like.

Injections can be prepared using a vehicle, such as saline solution, glucose solution, a mixture of both, or similar. Alternatively, powdered injections can be prepared by lyophilizing the antibody composition in the usual manner and adding sodium chloride.

Suppositories may be prepared using a vehicle such as cocoa butter, hydrogenated fat, carboxylic acid or the like.

In addition, aerosols can be prepared using the antibody composition as such or using a carrier that does not stimulate the patient's oral or respiratory mucous membrane and that can facilitate absorption of the antibody composition by dispersing it as fine particles.

Examples of carriers include lactose, glycerol, and the like. Depending on the properties of the antibody composition and the carrier, it is possible to produce pharmaceutical preparations such as aerosols, dry powders, and the like. Furthermore, the components exemplified as additives for oral preparations can also be added to parenteral preparations.

Although the clinical dose or frequency of administration varies depending on the target therapeutic effect, the method of administration, the treatment period, age, body weight and the like, it is normally 10 µg/kg to 20 mg/kg per day per adult.

It should be noted that the embodiments and features described in the context of one aspect of the present invention also apply to the other aspects of the invention.

Definitions

General glycobiology

The basic principles and definitions of glycobiology are described in Varki et al. Essentials of Glycobiology, 2nd edition, 2009.

"N-glycosylation" refers to the attachment of the sugar molecule oligosaccharide known as glycan to a nitrogen atom residue of a protein.

"O-glycosylation" refers to the attachment of a sugar molecule to an oxygen atom on an amino acid residue in a protein.

"Galactosylation" means the enzymatic addition of a galactose residue to lipids, carbohydrates, or proteins. "Sialylation" is the enzymatic addition of a neuraminic acid residue.

"Neuraminic acid" is a 9-carbon monosaccharide, a derivative of a ketone.

The "monoantennary" N-linked glycan is a designed N-glycan consisting of the N-glycan core (Manal-6(Mana1-3)Manp1-4GlcNAcp1-4GlcNAcp1-Asn-X-Ser/Thr) elongated with a single GlcNAc residue linked to C-2 and mannose A1-3. The single GlcNAc residue can be further elongated, for example, with Gal or Gal and NeuAc residues.

The "biantenary" N-linked glycan is the simplest of the complexes. N-linked glycans consist of an elongated N-glycan core (Mana1-6(Mana1-3)Manp1-4GlcNAcp1-4GlcNAcp1-Asn-X-Ser/Thr) with two GlcNAc residues linked to C-2 and mannose a1-3 and mannose a1-6. This core structure can then be elongated or modified by various glycan structures.

"Triantennary" N-linked glycans are formed when an additional GlcNAc residue is added to C-4 of the a1-3 core mannose or to C-6 of the a1-6 core mannose of the bi-antennary core structure. This structure can then be elongated or modified by a variety of glycan structures. "Tetraantennary" N-linked glycans are formed when two additional GlcNAc residues are added to C-4 of the a1-3 core mannose or to C-6 of the a1-6 core mannose of the bi-antennary core structure. This core structure can then be elongated or modified by a variety of glycan structures. "Poly-LacNAc" poly-N-acetyllactosamine ([Galp1-4GlcNAc]n; n > 2.)

"Glycoprofiling" means the characterization of the glycan structures resident in a biological molecule or cell.

The "glycosylation pathway" refers to the assembly of monosaccharides into a group of related complex carbohydrate structures by the stepwise action of enzymes, known as glycosyltransferases. Glycosylation pathways in mammalian cells are classified into N-linked protein glycosylation, different O-linked protein glycosylations (O-GalNAc, O-GlcNAc, O-Fuc, O-Glc, O-Xyl, O-Gal), different series of glycosphingolipids and GPI anchors.

"Biosynthetic step" means the addition of a monosaccharide to a glycan structure.

Glycosyltransferases are enzymes that catalyze the formation of the glycosidic bond to form a glycoside. These enzymes use activated sugar phosphates as glycosyl donors and catalyze the transfer of the glycosyl group to a nucleophilic group, usually an alcohol. The glycosyl transfer product can be an O-, N-, S-, or C-glycoside; the glycoside can be part of a monosaccharide, oligosaccharide, or polysaccharide.

"Glycosylation capacity" means the ability to produce a quantity of a specific glycan structure by a given cell or a given glycosylation process.

"HEP" is an abbreviation for heparosan. Heparosan (HEP) is a natural sugar polymer comprising (-GlcUA-1,4-GlcNAc-1,4-) repeats. It belongs to the glycosaminoglycan polysaccharide family and is a negatively charged polymer at physiological pH. It is found in the capsule of certain bacteria, but also in higher vertebrates, where it serves as a precursor to the natural polymers heparin and heparan sulfate.

"PEG" is an abbreviation for polyethylene glycol. PEG polymers are generally used to increase the half-life of therapeutic molecules. With their high hydrodynamic volume, PEG polymers increase the effective size of drugs and therefore delay their elimination from the bloodstream through renal filtration or by shielding the protein drug from elimination receptors and proteolytic degradation.

XTENs are well-defined peptide sequences based on a limited subset of amino acids and assembled into longer repeat sequences. XTEN also increases the size of therapeutic molecules and prolongs their presence in the bloodstream. GlycoPEGylation is the process of covalently attaching PEG to the glycans of a protein of interest.

"Enzymatic glycoPEGylation" is the process of covalently attaching PEG to the glycans of a protein of interest using a glycosyltransferase and suitable PEGylated glycosyl transfer groups, such as PEGylated NeuAc-CMP molecules.

"Glycoconjugate" is a macromolecule containing monosaccharides covalently linked to proteins or lipids.

"Simplest glycan structure" is a glycan structure that contains fewer monosaccharides and/or has lower mass and/or has fewer antennae.

"Human-like glycosylation" means having glycan structures similar to those of human cells. Examples include more α2,6-linked sialic acids (more α2,6 sialyltransferase enzyme) and/or fewer α2,3-linked sialic acids and/or more N-acetylneuraminic acid (Neu5Ac) and/or less N-glycolylneuraminic acid (Neu5Gc).

"Cleavage" refers to the breaking of the covalent backbone of a DNA molecule. Cleavage can be initiated by a variety of methods, including, but not limited to, enzymatic or chemical hydrolysis of a phosphodiester bond. Both single-strand and double-strand cleavage are possible, and double-strand cleavage can occur as a result of two distinct double-strand cleavage events. DNA cleavage can result in the production of blunt or staggered ends.

In certain embodiments, the fusion polypeptides are used for targeted cleavage of double-stranded DNA.

"Deconstruction" means obtaining cells that produce simpler glycan structures by single or stacked deletion of glycosyltransferases. Deconstruction of a glycosylation pathway means knocking out the glycosyltransferases involved in each step of biosynthesis and identifying the glycosyltransferases that control each biosynthetic step.

"Reconstruction" refers to the insertion of exogenous genes into cells or the activation of silent endogenous genes to obtain more complex glycan structures. Typically, the target cells produce simple glycan structures as a result of deconstruction.

"Modified glycan profile" refers to the change in the number, type, or position of oligosaccharides on the glycans of a given glycoprotein.

More "homogeneous glycosylation" or "homogeneous stable glycosylation capabilities" means that the proportion of identical glycan structures observed when glycoprofiling a given protein expressed in one cell is greater than the proportion of identical glycan structures observed when glycoprofiling the same protein expressed in another cell. More homogeneous glycosylation can be obtained by inserting and/or deleting one or more glycosyltransferase genes into a cell. An example of such a modification might be the deletion of B4galt1 to generate more homogeneous N-glycans without galactose. The further deletion of fut8 allows the production of more homogeneous N-glycans without galactose and fucose. In general, such knock-in and/or knockout modifications can result in changes of the resulting glycostructure of the modified cell such as the non-limiting list of simpler glycan structure, more homogeneous product, more sialic acids per molecule, non-sialylated, non-galactosylated, more homogeneous biantennary, more homogeneous monoantennary, more homogeneous triantennary, more homogeneous glycosylation without poly-LacNAc, higher productivity in cell culture, new stable homogeneous glycosylation capabilities, more human glycostructure, more homogeneous glycosylation and improved ADCC targeting of IgG, modified fucose level, no fucose and improved substrate for generating glycoconjugates.

General DNA, molecular biol. Any of various techniques used to separate and recombine DNA segments or genes, commonly by using a restriction enzyme to cut a DNA fragment from donor DNA and insert it into a plasmid or viral DNA. Using these techniques, DNA encoding a protein of interest is recombined/cloned (using PCR and/or restriction enzymes and DNA ligases or ligation-independent methods such as USER cloning) into a plasmid (known as an expression vector), which can subsequently be introduced into a cell by transfection using various transfection methods, such as calcium phosphate transfection, electroporation, microinjection, and liposome transfection. An overview, supporting information, and methods for constructing synthetic DNA sequences, insertion into plasmid vectors, and subsequent transfection into cells can be found in Ausubel et al., 2003 and/or Sambrook & Russell, 2001.

"Gene" refers to a region of DNA (including exons and introns) that encodes a gene product, as well as all regions of DNA that regulate the production of the gene product, whether such regulatory sequences are adjacent to coding and/or transcribed sequences or located remote from the gene whose function they regulate. Accordingly, a gene includes, but is not necessarily limited to, promoter sequences, terminators, translational regulatory sequences such as ribosome binding sites and internal ribosome entry sites, enhancers, silencers, insulators, delimiter elements, origins of replication, array binding sites, and locus control regions.

“Targeted genetic modifications”, “gene editing” or “genome editing”

Gene editing or genome editing refers to a process by which a specific chromosomal sequence is modified. The edited chromosomal sequence may comprise an insertion of at least one nucleotide, a deletion of at least one nucleotide, and/or a substitution of at least one nucleotide. In general, genome editing inserts, substitutes, or deletes nucleic acids from a genome using artificial nucleases such as zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), the CRISPR/Cas system, and meganuclease-reengineered homing endonucleases. The principles of genome editing are described in Steentoft et al., and gene editing methods are described in their references and are also widely used and, therefore, known to those skilled in the art.

"Endogenous" sequence/gene/protein refers to a chromosomal sequence, gene, or protein that is native to the cell or originates within the cell or organism being analyzed.

"Exogenous" sequence or gene refers to a chromosomal sequence that is not native to the cell, or a chromosomal sequence whose native chromosomal location is in a different location on a chromosome or originates outside the cell or organism being analyzed.

"Inactivated chromosomal sequence" refers to a genomic sequence that has been edited, resulting in the loss of function of a particular gene product. The gene is said to have been deleted.

"Heterologous" refers to an entity that is not native to the cell or species of interest.

The terms "nucleic acid" and "polynucleotide" refer to a polymer of deoxyribonucleotide or ribonucleotide, in linear or circular conformation, and in single-stranded or double-stranded form. For the purposes of this disclosure, these terms should not be construed as limiting the length of a polymer. The terms may encompass known analogs of naturally occurring nucleotides, as well as nucleotides that are modified at the base, sugar, and/or phosphate moieties (e.g., phosphorothioate structures). In general, an analog of a particular nucleotide has the same base-pairing specificity; that is, an analog of A will pair with T.

The term "nucleotide" refers to deoxyribonucleotides or ribonucleotides. Nucleotides can be standard nucleotides (i.e., adenosine, guanosine, cytidine, thymidine, and uridine) or nucleotide analogs. A nucleotide analog refers to a nucleotide that has a modified purine or pyrimidine base or a modified ribose moiety. A nucleotide analog can be a naturally occurring nucleotide (e.g., inosine) or a non-naturally occurring nucleotide.

The terms "polypeptide" and "protein" are used interchangeably to refer to a polymer of amino acid residues. These terms may also refer to glycosylated variants of "polypeptide" or "protein," also referred to as "glycoprotein," "polypeptide," "protein," and "glycoprotein" are used interchangeably throughout this disclosure.

The term "recombination" refers to a process of exchange of genetic information between two polynucleotides. For the purposes of this disclosure, "homologous recombination" refers to the specialized form of such exchange that occurs, for example, during the repair of double-strand breaks in cells. This process requires sequence similarity between the two polynucleotides, uses a "donor" or "exchange" molecule for the repair of the "target" molecule (i.e., the one that has suffered the double-strand break), and is known as "non-crossover gene conversion" or "short-tract gene conversion" because it results in the transfer of genetic information from the donor to the target. Without being bound by any particular theory, such transfer may involve the correction of mismatches in the heteroduplex DNA formed between the broken target and the donor, and/or "synthesis-dependent strand annealing," in which the donor is used to resynthesize the genetic information that will become part of the target, and/or related processes. Such specialized homologous recombination often results in an alteration of the target molecule's sequence such that part or all of the donor polynucleotide sequence is incorporated into the target polynucleotide.

As used herein, the terms "target site" or "target sequence" refer to a nucleic acid sequence that defines a portion of a chromosomal sequence to be edited and to which a targeting endonuclease is designed to recognize, bind, and cleave.

"Targeted integration" is the method by which exogenous nucleic acid elements are specifically integrated into defined loci of the cell's genome. Specific double-strand breaks are introduced into the genome by genome-editing nucleases, which allow the integration of the exogenously administered donor nucleic acid element into the double-strand break site. In this way, the exogenously supplied donor nucleic acid element is stably integrated into the defined locus of the cell's genome.

Sequence identity techniques for determining the identity of nucleic acid and amino acid sequences are known in the art. Typically, such techniques include determining the nucleotide sequence of a gene's mRNA and/or determining the amino acid sequence encoded by it, and comparing these sequences with a second nucleotide or amino acid sequence. Genomic sequences can also be determined and compared in this manner. In general, identity refers to an exact nucleotide-to-nucleotide or amino acid-to-amino acid match between two polynucleotides or polypeptide sequences, respectively. Two or more sequences (polynucleotides or amino acids) can be compared by determining their percentage identity. The percent identity of two sequences, whether nucleic acid or amino acid sequences, is the number of exact matches between two aligned sequences divided by the length of the shorter sequences and multiplied by 100. The local homology algorithm of Smith and Waterman, Advances in Applied Mathematics 2:482–489 (1981) provides an approximate alignment of nucleic acid sequences. This algorithm can be applied to amino acid sequences using the scoring matrix developed by Dayhoff, Atlas of Protein Sequences and Structure, M.O. Dayhoff ed., 5 suppl. 3:353–358, National Biomedical Research Foundation,

Washington, DC, USA, and normalized by Gribskov, Nucl. Acids Res. 14(6):6745–6763 (1986). An exemplary implementation of this algorithm for determining percent sequence identity is provided by the Genetics Computer Group (Madison, Wis.) in the application program "BestFit". Other suitable programs for calculating percent sequence identity or similarity are generally known in the art; for example, another alignment program is BLAST, used with default parameters. For example, BLASTN and BLASTP can be used with the following default parameters: genetic code=std; filter=none; string=both; cutoff=60; expect=10; matrix=BLOSUM62; descriptions=50 sequences; sort by=HIGHSCORE; databases=nonredundant, GenBank+EMBL+DDBJ+PDB+GenBank CDS translations+Swiss protein+Spupdate+PIR. Detailed information about these programs can be found on the GenBank website. For the sequences described herein, the desired degree of sequence identity ranges from approximately 80% to 100%, and any integer value in between. Typically, the percentages of identity between sequences are at least 70-75%, preferably 80-82%, more preferably 85-90%, even more preferably 92%, still more preferably 95%, and most preferably 98% sequence identity.

All patent and non-patent references cited in this application are incorporated by reference in their entirety.

The invention will now be described in more detail in the following non-limiting examples.

Examples

The following examples are intended to illustrate various embodiments of the invention and are therefore not intended to limit the present invention in any way. Together with the present examples, the methods described herein are presently representative of preferred embodiments; they are exemplary and are not intended to limit the scope of the invention. Changes to the invention and other uses that are within the spirit of the invention as defined by the scope of the claims will occur to those skilled in the art.

Example 1: Deconstruction of n-glycosylation in cho cells

Introductory paragraph

The Chinese hamster ovary (CHO) cell is the cell factory of choice for the recombinant production of biologic therapies. CHO is capable of many post-translational modifications required for the production of bioactive proteins and has a generic capacity for protein glycosylation compatible with human use. Glycosylation has a wide range of effects on the production, product consistency, solubility, stability, bioactivity, and circulatory half-life of protein therapies, and the availability of a single generic CHO production platform is a limiting factor for the design and development of biologics. This example shows a comprehensive gene knockout screen for glycosyltransferases involved in N-glycosylation in CHO, defining the key enzymes controlling antennal branching, elongation, and sialylation. A panel of stable engineered CHO lines with designer glycosylation capabilities was established, including homogeneous biantennary glycans with and without α2,3- and α2,6-sialic acid coatings. The screen demonstrates high plasticity for CHO glycoengineering and provides clear avenues for custom designs.

Introduction

Glycoprotein-based biologics are the fastest-growing therapeutic class, and most can only be produced recombinantly in mammalian cells with glycosylation capacity similar to that of humans. The Chinese hamster ovary (CHO) cell has acquired a prominent role as a host cell for the recombinant production of therapeutic glycoproteins, primarily because it produces fairly simple N-glycans with branching and capping similar to those produced in some human cells.

In particular, CHOs produce heterogeneous complex-type N-glycans with bi-, tri- and tetra-antennary structures with an α6-fucose (Fuc) core, a minor amount of poly-N-acetyllactosamine (poly-LacNAc) mainly on the α1,6 arm of the tetra-antennary structures, and a unique LacNAc coating with α2,3-linked neuraminic acid (NeuAc) (see Fig. 1a). CHO generally does not produce the immunogenic non-human and human coating structures N-glycolylneuraminic acid (NeuGc) or α3Gal, although their appearance has been reported, perhaps as a result of gene induction. N-glycosylation can vary for different proteins, as well as for different glycosites within individual proteins. For example, IgG antibodies are produced with truncated N-glycan structures at the conserved glycosite Asn297 (biantenary structures with a6Fuc core, limited LacNAc, and NeuAc cap). A major concern in CHO is the substantial heterogeneity in N-glycan processing, which can be difficult to control during bioprocessing and pose bioactivity and biosafety issues. Thus, a major focus of therapeutic product bioprocessing is on glycan analysis and fermentation control to achieve consistency.

In the last two decades, significant efforts have been devoted to the genetic glycoengineering of CHO cells with the aim of expanding glycosylation capacity, reducing heterogeneity, and specifically improving or altering sialylation.

All of these studies have essentially used random integration of cDNAs encoding glycosyltransferases and have experienced problems with stability, consistency, and predictability of the introduced glycosylation capacity. The main obstacle has been the need to rely on overexpression of glycosyltransferases and competition with endogenously expressed enzymes due to the lack of straightforward methods to knock them down in cell lines. Therefore, to our knowledge, these types of glycoengineered CHO cells have not progressed to clinical therapeutics. However, a successful glycoengineering strategy has emerged following the discovery that coreless IgGs targeting 6Fuc at the N-glycan Asn297 exhibit significantly superior antibody-dependent cellular cytotoxicity (ADCC).

Thus, by a tour-de-force using two rounds of homologous recombination, both alleles of the fut8 gene encoding the α6-fucosyltransferase that controls core fucosylation were deleted in CHO, and at least one therapeutic IgG produced in CHO lacking the fut8 gene is now in clinical use. More recently, the fut8 gene was deleted by precise gene editing with zinc finger nuclease (ZFN) with a fraction of the time and resources invested. The emergence of precise gene editing technologies for knockout (KO) and knockin (KI) has opened the door to an entirely different level of speed and ease with which stable genetic manipulation of host cell lines to delete and introduce glycosyltransferase genes can be achieved, which will undoubtedly impact the engineering of mammalian host cell factories for the recombinant production of therapeutics.

In this case, we employed for the first time a ZFN-mediated knockout screen in CHO to explore the potential for engineering N-glycosylation of recombinant glycoproteins. Many steps in the N-glycan biosynthetic pathway are potentially catalyzed by multiple isozymes, making genetic engineering unpredictable (Fig. 1a). The in vivo functions of each of these isozymes need to be analyzed to construct a matrix of design options. The knockout screen included all genes encoding isozymes with potential to regulate N-glycosylation branching, elongation, and end-capping expressed in CHO. Expression of relevant genes in CHO GS was assessed by quantitative next-generation RNA sequencing (RNAseq) (Figure 4). This strategy was only possible thanks to the recent sequencing of the CHO genome and analysis of the CHO-K1 transcriptome. We focus on individual genes and the most relevant stacked combinations.

We used a total of 15 ZFNs targeting glycosyltransferase genes (Fig. 1 a) involved in antennal branching (mgat4A/4B/4C/5/5B) (Fig. 1 c), galactosylation (B4galt1/2/3/4) (Fig. 1 d), poly-LacNAc elongation (B3gnt1/2/8) (Fig. 1 e), and end-capping by sialylation (st3gal3/4/6) (Fig. 1 f). We used recently developed methods to enrich for KO clones by FACS (GFP/Crimson-tagged ZFNs) and high-throughput screening using an amplicon labeling strategy (IDAA). The majority of KO clones had insertions and/or deletions (indels) in the range of ±20 bps, and most of the target genes were present with two alleles, while some (mgat4B and mgat5) were present with 1 or 3 alleles, respectively. To monitor the effects of glycogene knockout screening, we used stable recombinant expression of human erythropoietin (EPO), as this protein has three N-glycans with mainly tetraantennary structures, low poly-LacNAc, and an α2,3-sialic acid cap (Fig. 1b), and is one of the best-characterized N-glycoproteins produced in CHO. Engineering was performed in the CHO-GS production line (Sigma) grown in suspension in protein-free medium, and EPO expressed in the wild-type cell was found to be glycosylated similarly to what has been reported in the past with other CHO lines (Fig. 1b).

No consistent morphological or phenotypic growth characteristics were observed, apart from subtle variations within clones. On average, 5–50 clones of each were screened by ELISA, and the highest expressing clones were selected. Final purification yields ranged from 0.2 mg/L to 6.2 mg/L, but this variation was mainly, if not entirely, due to chance and the limited number of clones screened for each KO clone. The selected WT CHO clone produced final yields of 0.9 mg/L, whereas, for example, the KO clone mgat4A/4B/5/B3gnt2 with four genes deleted produced 3.4 mg/L.

Monitoring the status of N-glycan antennas

MGAT1 and 2 each control the formation of one of the two p2 branches in biantennary N-glycans, whereas a possible partial functional redundancy is predicted for the formation of tri- and tetraantennary branches by MGAT4A/4B/4C (p4 branch) and MGAT5/5B (p6 branch), respectively. Only mgat4B and 5 appear to be expressed in CHO (Fig. 4). Targeting mgat4A had a minor effect, whereas targeting mgat4A/4B almost abolished p4 branched tetraantennary N-glycans (Fig. 1 c). In contrast, targeting mgat5 completely abolished p6 branched tetraantennary structures (Fig. 1 c) as well as L-PHA lectin labeling of cells (Fig. 5), consistent with previous studies of lectin-resistant CHO mutants lacking mgat5. The KO stacking of mgat4A/4B and 5 produced homogeneous biantennary N-glycans in EPO (Fig. 1c). However, a lower amount of poly-LacNAc was found in the biantennary N-glycans, which is also present as a minor component in wild-type cells.

LacNAc control

CHOs express all seven known β4-galactosyltransferases, and our understanding of the in vivo functions of these isozymes is poor, but only B4Gal-T1-4 is expected to play roles in N-glycosylation. We first examined individual KO clones for B4galt1-4 and found that B4Gal-T1 appeared to have the major role (Fig. 1d). Thus, the contributions of the other isoforms could only be assessed in stacked combinations with B4galt1 KO. We tested the stacked combinations by immunocytology with an antibody against LacNAc, showing that only the stacked KO of B4galt1/3, but not B4galt1/2/4, appeared to abolish galactosylation (Supplementary Fig. 3). In agreement with this, we found that EPO expressed in stacked B4galt1/3 but not B4galt1/2 KO clones had a substantial reduction (>90%) in N-glycan galactosylation (Fig. 1d). However, complete loss of EPO galactosylation capacity may require the knockout of three or more genes.

Poly-LacNAc control

While poly-LacNAc on N-glycans can confer superior circulatory half-life properties to, for example, EPO, this modification is always incomplete and one of the most heterogeneous and difficult to control for product consistency. CHO cells generally produce low amounts of poly-LacNAc on N-glycans and mainly on tetraantennary structures on the p6 arm controlled by MGAT5. The biosynthesis of poly-LacNAc on N-glycans is poorly understood and three genes may be involved, B3gnt1, B3gnt2 and B3gnt8. We evaluated single gene knockout of B3gnt1, B3gnt2, and B3gnt8 with lectin immunocytology using the poly-LacNAc lectin LEL, and identified B3gnt2 as the key gene controlling LacNAc initiation in CHO cells (Fig. 5B). In agreement with this, EPO expressed in CHO clones with KO of B3gnt2 and not B3gnt1, resulted in the elimination of the minor fraction of poly-LacNAc in EPO (Fig. 1e).

Control of sialylation

CHOs produce almost exclusively NeuAc protection of N-glycans linked by α2,3 linkages. All six known ST3Gal transferases are expressed in CHO cells (Fig. 4), and the roles of each in N-glycan sialylation are poorly understood, but candidates are primarily the ST3Gal-III, ST3Gal-IV, and ST3Gal-VI isoforms. We first targeted these three genes individually using immunocytology to assess LacNAc exposure and found that only st3gal4 knockout produced substantial effects (Fig. 7). Analysis of EPO expressed in single and stacked KO clones showed that loss of st3gal4 and 6 produced substantial effects, whereas we found two patterns in st3gal4/6 knockout clones. In the WT background, KO of st3gal4/6 produced EPO with heterogeneous tetraantennary N-glycans without traces of sialylation (Fig. 1f). KO of st3gal4/6 in combination with mgat4A/4B/5 produced biantennary N-glycans without sialylation but with increased poly-LacNAc (Fig. 2b), or in one clone without increased poly-LacNAc but with smaller amounts of a mono-sialylated biantennary N-glycan. Therefore, targeting st3gal3/4/6 may be necessary to achieve a complete absence of sialic acid. Despite this, CHO clones without st3gal4 and 6 produced EPO with uncapped LacNAc termini, opening the door to de novo engineering of recombinant α2,6NeuAc-capped glycoproteins.

Custom glycoengineering of CHO to produce α2,6NeuAc-protected N-glycans

Most circulating glycoproteins are capped with α2,6NeuAc, whereas virtually all recombinant therapeutics are produced with α2,3NeuAc. There has long been interest in producing CHO cells capable of producing homogeneous N-glycans with α2,6NeuAc protection. To demonstrate the de novo glycoengineering prospects provided by our KO assay, we generated CHO clones capable of producing homogeneous α2,6NeuAc protection (Fig. 2). To circumvent previous problems with random plasmid integration, we used the ZFN-directed knockin (KI) of ST6Gal-I with a modified Obligare strategy at a SafeHarbor integration site. Human ST6GAL1 introduced into st3gal4/6 knockout CHOs produced the full range of antennal N-glycan structures with a normal degree of NeuAc coating (Fig. 2a), and when introduced into CHOs with an additional knockout of mgat4A/4B/5, EPO was produced with homogeneous biantennary N-glycans coated with α2,6NeuAc (Fig. 2b). Interestingly, de novo introduction of ST6GAL1 also abrogated the lower amounts of poly-LacNAc formed in biantennary structures when sialylation was abolished (Fig. 2b), demonstrating that the B3gnt2 knockout was also unnecessary. Thus, we were able to introduce the ability to sialylate α2,6NeuAc and produce a CHO cell capable of producing N-glycoproteins with a homogeneous α2,6NeuAc coating. The Obligare KI strategy was highly efficient with approximately 30–50% of individual cloned cells expressing ST6Gal-I as assessed by antibody and lectin immunocytology (Fig. 7), whereas relying on ZFN-mediated homologous recombination did not result in successful targeted integration in CHO cells. ST6GAL1 KI clones were stable for over 20 passages as assessed by SNA lectin immunocytology and MAb 1B2 (not shown).

In summary, the KO assay identified key glycogens that control critical steps in protein N-glycosylation in CHO (Fig. 3) and demonstrated remarkable plasticity in glycosylation tolerance with no evidence of unexpected compensatory changes in glycosylation. We provided design strategies for generating CHO host cells with homogeneous biantennary N-glycans with and without sialylation that would serve as platforms for de novo engineering of desired glycosylation capabilities. The array provides a general guideline for custom engineering CHO N-glycosylation capabilities that can likely be used for any CHO host line and even for established production lines. We demonstrated the feasibility of de novo engineering homogeneous glycosylation capabilities by introducing α2,6-sialylation into CHO using a precise KI strategy in CHO lines knocked out of endogenous α2,3-sialyltransferase genes. The generated CHO cell panel will enable de novo dissection and design of N-glycosylation for any protein therapy, and the production of a wide range of therapeutic protein glycoforms will allow for straightforward evaluation and selection of the optimal design. The advent of precise genetic engineering clearly promises a new era for today's largest producer of glycoprotein therapies: the CHO cell.

Methods

Targeting of ZFNs to glycogen in CHO cells. All gene targeting was performed either in the CHOZN GS-/- (glutamine synthase) clone produced by ZFN KO and provided by Sigma-Aldrich, St. Louis, MO, or in CHO-K1 obtained from ATCC. All CHO media, supplements, and other reagents used were obtained from Sigma-Aldrich unless otherwise specified. CHO GS cells were maintained as suspension cultures in EX-CELL CHO CD Fusion serum-free medium supplemented with 4 mM L-glutamine. Cells were seeded at 0.5 x 106 cells/mL in a T25 flask (NUNC, Denmark) one day before transfection. 2 x 106 cells and 2 μg of endotoxin-free plasmid DNA of each ZFN (Sigma, USA) were used for transfection. Each ZFN was labeled with GFP and Crimson using a 2A linker. Transfections were performed by electroporation using the Amaxa V kit and the U24 program with Amaxa Nucleofector 2B (Lonza, Switzerland). Subsequently, electroporated cells were plated in 3 mL of growth medium in a 6-well plate. Cells were transferred to 30°C for a 24-hour cold shock. 72 hours after nucleofection, the pool of cells with the 10–15% most labeled for both GFP and Crimson was enriched by FACS. Cells were re-sorted by FACS one week later to obtain individual clones in round-bottom 96-well plates. KO clones were identified by the recently described insertion–deletion analysis (IDAA) and, when possible, by immunocytology with appropriate lectins and monoclonal antibodies. The selected clones were further verified by TOPO for detailed characterization of the introduced mutations. This strategy allowed for rapid selection and detection of KO clones with frameshift mutations, and, on average, we selected between 2 and 5 clones from each target event. RNA seq analysis of the CHO clones was performed by BGI.

Recombinant expression of human EPO in CHO cells. An expression construct containing the complete coding sequence of human EPO cloned into pcDNA3.1/myc-His (C-terminal tags) was synthesized by Genewiz, USA. EPO is a 166 amino acid protein with N-glycans at Asn24, Asn38, and Asn83. CHO cells were co-transfected with EPO and GFP plasmids, followed by FACS to enrich for GFP-expressing clones. Stable transfectants were selected in 0.4 mg/mL Zeocin (Invitrogen). Positive stable clones were examined by direct enzyme-linked immunosorbent assays (ELISAs) using the anti-His(C-Term)-HRP monoclonal antibody (Invitrogen). His-tagged recombinant human EPO was purified by nickel affinity purification (Invitrogen, USA). The media were mixed 3:1 (v/v) in 4x binding buffer (200 mM Tris, pH 8.0, 1.2 M NaCl) and applied to 0.3 ml of packed NiNTA agarose (Invitrogen), pre-equilibrated in binding buffer (50 mM Tris, pH 8.0, 300 mM NaCl). The column was washed with binding buffer, and then the bound protein was eluted with binding buffer containing an additional 250 mM imidazole. EPO-containing fractions were determined by SDS-PAGE and further purified in a reverse-phase HPLC purification with a Jupiter C4 column (5 gm, 300 A, 250 x 4.6 mm column) (Phenomenex), using 0.1% trifluoroacetic acid (TFA) and a 10-100% acetonitrile gradient. Purity was assessed by Coomassie staining of SDS-PAGE gels and proteins were quantified using the BCA Protein Assay Kit (Thermo Scientific).

N-glycan profiling of recombinant EPO in CHO clones. Approximately 25 μg of purified EPO was digested with 2 U of PNGase F (Roche Diagnostics, Mannheim, Germany) in 50 mM ammonium bicarbonate at 37°C overnight. The released N-glycans were cleaved from rhEPO using in-house packed Stagetips (Empore disk-C18, 3M) and incubated in 100 mM ammonium acetate, pH 5.0 for 60 min at 22°C. N-glycans were dried in glass vials and permethylated as described above. Permethylated N-glycans were extracted with chloroform and desalted by washing the organic phase repeatedly with deionized water. The organic phase was evaporated under a stream of nitrogen gas and the permethylated N-glycans were dissolved in 20 gL of 50% methanol. Finally, 1 gL of sample was co-crystallized with 1 gL of matrix composed of 2,5-dihydroxybenzoic acid (10 mg/mL) in 70% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid and 2.5 mM sodium acetate. The permethylated N-glycans were analyzed by MALDI-TOF positive mode (Autoflex Speed, BrukerDaltronics, Bremen) operated in reflector mode with data acquisition in the m/z mass range of 1000 to 7000.

Example 2

Determination of the repertoire of glycosyltransferases expressed in a mammalian CHO cell line.

CHO-K1 was obtained from ATCC, and CHO-GS (clone CHOZN GS-/- (glutamine synthase) produced by ZFN KO) was obtained from Sigma-Aldrich, St. Louis, MO. All CHO media, supplements, and other reagents used were obtained from Sigma-Aldrich unless otherwise specified. CHO-GS cells were maintained as suspension cultures in serum- and animal component-free EX-CELL CHO CD Fusion media supplemented with 4 mM L-glutamine. CHO cells were seeded at 0.25x106cells/mL in a 6-well plate and harvested in exponential phase 48 h postinoculation for total RNA extraction using the RNeasy minikit (Qiagen). RNA integrity and quality were checked using the 2100 Bioanalyzer (Agilent Technologies). Library construction and next-generation sequencing were performed using the Illumina HiSeq 2000 system (Illumina, USA) under standard conditions recommended by the ARNsec service provider. The aligned data were used to calculate the distribution of reads across CHO reference genes, and coverage analysis was performed. Only alignment results that passed quality control were used for subsequent gene expression analysis.

RNAseq analysis was performed in two common CHO lines (CHO-K1, CHO-GS) and two independent triple mgat4a/4b/5KO CHO-GS clones (ZFN91-1C8, ZFN91-2A6) (TABLE 3). The reported RNAseq analysis of CHO-K1 was included as a reference and was largely similar to our analyses, except that relative expression levels were higher. Notably, some genes, such as galntly mgat4b, that were not expressed in CHO-K1 (Xu, Nagarajan et al. 2011), were now expressed. Furthermore, in the case of mgat4b, this gene was found to be essential for the glycoengineering experiments carried out here.

An important observation from the analysis of the CHO transcriptomes was that the glycogen expression profiles in the analyses of the two triple KO clones were identical to those of the parental CHO lines, demonstrating that targeted KO gene editing performed in the CHO cell lines did not alter the expression of other glycosyltransferase genes.

Example 3

Genetic inactivation of glycosyltransferase genes in the N-glycosylation pathway (Deconstruction).

All targeted glycosyltransferase gene knockouts were performed in CHO-GS and/or CHO-K1 and cells were cultured as described in Example 2. Cells were seeded at 0.5 x 106 cells/mL in a T25 flask (NUNC, Denmark) one day prior to transfection. 2 x 106 cells and 2 pg of endotoxin-free plasmid DNA of each ZFN (Sigma, USA) were used for transfection. Each ZFN was labeled with GFP and Crimson using a 2A linker (Duda, Lonowski et al. 2014). Transfections were performed by electroporation using the Amaxa V kit and the U24 program with Amaxa Nucleofector 2B (Lonza, Switzerland). Subsequently, electroporated cells were seeded in 3 mL of culture medium in a 6-well plate. Cells were moved to 30°C for a 24-hour cold shock. 72 hours after nucleofection, the top 10–15% of GFP- and Crimson-labeled cell pool were enriched by FACS. We used recently developed methods to enrich for KO clones by FACS (GFP/Crimson-labeled ZFNs) (Duda, Lonowski et al. 2014) and a high-throughput detection by amplicon labeling (IDAA) strategy (Zhang, Steentoft et al. submitted). Cells were re-sorted by FACS 1 week later to obtain individual clones in round-bottom 96-well plates. KO clones were identified by IDAA and, where possible, also by immunocytology with appropriate lectins and monoclonal antibodies. The selected clones were subsequently verified by TOPO cloning of PCR products (Invitrogen, USA) for detailed characterization by Sanger sequencing of the introduced mutations in individual TOPO clones. This strategy allowed rapid selection and detection of KO clones with appropriate inactivating mutations as described here, and on average we selected between 2 and 5 clones from each target event.

Most KO clones had out-of-frame insertions and/or deletions (indels) in the range of ±20 bps, and most target genes were present with two alleles, while some (mgat4B and mgat5) were present with 1 or 3 alleles, respectively (TABLE 4).

TABLE 4 lists all individual and stacked glycosyltransferase gene inactivation events and the order (line of ancestry) of stacking of glycosyltransferase gene inactivation in cells to produce a deconstruction matrix of the N-glycosylation pathway.

Example 4

Insertion of glycosyltransferase genes into the N-glycosylation pathway (Reconstruction).

Target-specific integration (knockin/KI) was directed toward the CHO Safe Harbor #1 locus (S. Bahr et al., BMC Proceedings 2013, 7(Suppl 6):P3. doi:10.1186/1753-6561-7-S6-P3). A modified ObLiGaRe strategy (33) was used where two inverted ZFN binding sites flank the donor plasmid gene of interest to be inserted. First, a shuttle vector designated ObLiGaRe-2X-Ins was synthesized (TABLE 5). ObLiGaRe-2X-Ins was designed such that any cDNA sequence encoding a protein of interest can be inserted into a multiple cloning site where transcription is driven by the CMV IE promoter and a BgH terminator. To minimize epigenetic silencing, two insulator elements flanking the transcription unit were inserted. Additionally, an AAVS1 “landing platform” was included at the 3’ end of ObLiGaRe-2X-Ins just upstream of the 3’ inverted ZFN binding site. A complete ST6GAL1 open reading frame was directionally inserted into ObLiGaRe-2X-Ins generating ObLiGaRe-2X-Ins-ST6GAL1 (TABLE 5). Transfection and clone sorting were performed as described in EXAMPLES 2 and 3. Clones were initially screened by positive SNA lectin staining and selected clones were subsequently analyzed by 5’ and 3’ junction PCR to confirm correct targeted integration into the Safeharbor#1 site of the CHO.st3gal4/6KO cell. The allelic copy number of integration was determined by WT allelic PCR. Subsequently, the full-length human MGAT4A open reading frame was directionally inserted into the second allele of Safe Harbor#1 in the ST6GAL1 KI clone, followed by insertion of full-length human MGAT5 into the AAVS1 “landing platform,” which was included at the 3' end of ObLiGaRe-2X-Ins just upstream of the 3' inverted ZFN binding site (TABLE 5).

Example 5

Recombinant production of human EPO and IgG in gene-edited mammalian CHO cells.

Expression constructs containing the complete coding sequence of human EPO and a therapeutic IgG cloned into pcDNA3.1/myc-His (C-terminal tags) and pBUDCE4.1 (Invitrogen), respectively, were synthesized by Genewiz, USA. EPO is a 166 amino acid protein with N-glycans at Asn24, Asn38 and Asn83, and IgG has a single N-glycan at Asn297, respectively. Wt or KO CHO cells were transfected with EPO or IgG and maintained at 37 °C as suspension cultures in serum-free EX-CELL CHO CD Fusion medium supplemented with 4 mM L-glutamine and 400 μg/mL Zeocin in a 5% CO2 incubator. Clones were expanded and grown in T75 flasks at a seeding density of 0.5x106 cells/mL and harvested 72 or 96 hours later. Cell viability was greater than 90% in all reporter-expressing clones, and the viable cell density was between 2 and 3x106 cells/mL at the time of harvest. Cells were removed by centrifugation at 300 g for 5 minutes, and the supernatant was stored at -80°C. His-tagged recombinant human EPO was purified by nickel affinity purification (Invitrogen, USA). The media were mixed 3:1 (v/v) in 4x binding buffer (200 mM Tris, pH 8.0, 1.2 M NaCl) and applied to 0.3 mL of packed NiNTA agarose (Invitrogen), pre-equilibrated in binding buffer (50 mM Tris, pH 8.0, 300 mM NaCl). The column was washed with binding buffer, and then the bound protein was eluted with binding buffer containing an additional 250 mM imidazole. EPO-containing fractions were determined by SDS-PAGE and further purified in a reverse-phase HPLC purification with a Jupiter C4 column (5 gm, 300 A, 250 x 4.6 mm column) (Phenomenex), using 0.1% trifluoroacetic acid (TFA) and a 10-100% acetonitrile gradient. IgG was purified by HiTrap™ Protein G Hp (GE Healthcare, USA) pre-equilibrated/washed in PBS and eluted with 0.1 M glycine (pH 2.7). Protein purity was assessed by Coomassie staining of SDS-PAGE gels, and proteins were quantified using the BCA Protein Assay Kit (Thermo Scientific, Rockford, USA).

Example 6

N-glycan profile of purified EPO and IgG.

Approximately 25 μg of purified EPO was digested with 2 U of PNGase F (Roche Diagnostics, Mannheim, Germany) in 50 mM ammonium bicarbonate at 37°C overnight. For IgG, 35 μg of samples were initially digested (overnight, 37°C) with trypsin in 50 mM ammonium bicarbonate, then heated at 95°C for 15 min and cooled to RT before PNGase F digestion as described above. The released N-glycans were separated from intact EPO or the IgG tryptic digest using in-house packaged Stagetips (Empore-C18 disc, 3M) and incubated in 100 mM ammonium acetate, pH 5.0 for 60 min at 22°C. The N-glycans were divided into three equal aliquots. One third was dried in glass vials and permethylated as described previously (Ciucanu and Costello 2003) and the remainder was saved for native N-glycan analysis. Permethylated N-glycans were extracted with chloroform and desalted by washing the organic phase repeatedly with deionized water. The organic phase was evaporated under a stream of nitrogen gas and the permethylated N-glycans were dissolved in 20 gL of 50% (v/v) methanol. Finally, 1 gL of sample was co-crystallized with 1 gL of matrix composed of 2,5-dihydroxybenzoic acid (10 mg/mL) in 70% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid and 2.5 mM sodium acetate. Permethylated N-glycans were analyzed by MALDI-TOF in positive mode (Autoflex Speed, BrukerDaltronics, Bremen) operated in reflector mode with data acquisition (2000 shots/spot) in the mass range m/z 1000–7000. Glycan composition analysis of all spectra was performed using the SysBioWare platform and is presented in Supplementary Table 4 (Vakhrushev, Dadimov et al. 2009). The released native N-glycans were reconstituted in MeOH/HzO (1/1; v/v, containing 50 mM ammonium bicarbonate) at a concentration of 0.5 pmol/gL and analyzed on an OrbiTrap Fusion MS (Thermo San Hose, USA). Samples were infused directly through EASY-Spray and analyzed by negative ion mode. The flow rate was set at 500 nL/min, the spray voltage at 1900–2100 V, and the ion transfer tube temperature at 275 °C. Precursor ions (isolation width, m/z 3) were selected and subjected to HCD fragmentation at 35% normalized collision energy (default charge state, z=2). All spectra were recorded with a resolution of 120,000.

Example 7

Inactivation of glycosyltransferase genes by precise multi-exonic gene editing with tandem single-stranded oligodeoxynucleotide (ssODN)/ZFN.

In this example, gene inactivation is achieved by deleting the exons encoding the N-terminal cytoplasmic tail, the transmembrane region, and parts of the stem region of the target glycosyltransferase gene. Deletion of the target region is mediated by using the tandem single-stranded oligodeoxyribonucleotide (ssODN)/ZFN approach (Duda, Lonowski et al. 2014). The ZFN target site is selected within or as close as possible to the sequence encoding the transmembrane region of the glycosyltransferase. As an example, CHOmgat4 is selected using the ComposR ZFN from CHOmgat4a (Sigma) and an ssODN possessing 60bp homology arm sequences flanking the 5' UTR region, including the translation start site and the 3'-flanking exon 2 ZFN target site of mgat4a (ssODN mgat4a:gagagagattggtcttgcttgtcatcaccaacgtatgaaccagtgtgatggtgaaatgagtcGCCGTGCAGGA GCAGCAACTAACGGAAGTAACACTGCATTGACTACATTTTCAGGTAC) (the 5' UTR region is shown in lowercase, the partial ZFN cleavage site is in italics, and the 3'-flanking exon 2 ZFN junction site is in uppercase), thus removing the cytoplasmic tail encoding exon 1-2. the transmembrane domain and the stem of mgat4a. Gene knockdown was performed in CHO-GS as described in Example 2, including mgat4a ODN. Transfections were performed by electroporation using the Amaxa V kit and the U24 program with Amaxa Nucleofector 2B (Lonza, Switzerland). Subsequently, electroporated cells were plated in 3 mL of growth medium in a 6-well plate. Cells were moved to 30°C for a 24-hour cold shock. 72 h after nucleofection, 10-15% of the most GFP- and Crimson-labeled cell pool were enriched by FACS. Cells were re-sorted by FACS one week later to obtain individual clones in round-bottom 96-well plates. Enrichment of KO clones was performed by FACS (GFP/Crimson-labeled ZFNs). Clones were analyzed by PCR using primers Mgataex1 F (5'-TATCCACTGTGTTGCTTGCTG-3')/Mgataex2R (5'-Actgctcttccagaggtcctg d-3'), detecting only correctly deleted clones. Mono/biallelic orientation was determined by detection of the presence of wt alleles by PCR using primers Mgataex2F (5'-gaacgccttcgaatagctgaacatagg-3')/Mgataex2R. Genetic PCR results were validated by Southern blot analysis using an intron 2-specific probe that detects correct deletion of the 37 KBp intron 1 target region containing diagnostic KpnI sites.

Example 8

Inactivation of glycosyltransferase genes by dual CRISPR/Cas9 multi-exon gene editing.

Genetic inactivation is ensured by removing the exons encoding the cytoplasmic/transmembrane tail and parts of the stem coding sequences of the desired gene that needs to be inactivated. The elimination of the targeted region is mediated by the use of a CRISPR/Cas9-gRNA pair targeting the flanking regions encoding the Mgat4a signal sequence, the transmembrane anchor, and the stem regions. CHOMgat4a is targeted using Mgat4aEx1gRNA from CHOMgat4aCRISPR/Cas9 gRNA (5'-GGTATACCACATGGCAAAATGGG-3') specific for exon 1 and Mgat4aEx2gRNA (5'-GTCCAACAGTTTCGCCGTGCAGG-3') specific for exon 2. Both gRNAs were cloned into the BbsI (NEB, USA) gRNA target site of px458 (Addgene, USA) encoding the GFP-tagged Cas9 of S. pyogenes and the U6 promoter-driven gRNA cassette, generating px458-CRISPR-Mgat4aexlgRNA and px458-CRISPR-Mgat4aex2gRNA. Precise gene editing was performed by nucleofection using 2 ug of these CRISPR/Cas9-Mgat4a editing tools and targeting validation in CHO-GS cells as described in Examples 3 and 7.

Example 9

Enzymatic glycoPEGylation of single-antenna N-glycans. KO of mgat2 in combination with mgat4a/4b/5 in CHO cells resulted in homogeneous single-antenna N-glycans when EPO was expressed (FIG. 32). Reintroduction of human MGAT4A as well as MGAT4A and 5 in combination by knockin resulted in tri- and tetra-antenna N-glycans, respectively (FIG. 32), demonstrating that the obtained single-antenna N-glycan is linked (31-2 to the a1-3Man branch controlled by mgatl. Enzymatic glycoPEGylation or transfer of other compounds by sialyltransferases and CMP-(PEG)NeuAc or a relevant modified donor was tested using EPO as an example (FIG. 33-34).

Material and methods EPO-Myc-His6 glycovariants were produced in CHO with KOmgat2/4a/4b/5 and purified on a nickel-NTA column and buffer exchanged into 100 mM Tris, 150 mM NaCl, 0.005% NaN3, pH 7.2 buffer as described in Example 4. Wild-type EPO-Myc-His6 with normal glycosylation profile was produced in unmodified CHO cells and similarly purified on a nickel-NTA column before buffer exchange into 100 mM Tris, 150 mM NaCl, 0.005% NaN3, pH 7.2 buffer. Protein concentrations were determined using the NanoDrop Lite Spectrophotometer (Thermo Scientific™). Sialyltransferase (ST3Gal-III; 1.35 U/mL) in 20 mM HEPES, 120 mM NaCl, 50% glycerol, pH 7.0 was obtained from FujiFilm Diosynth Biotechnologies, Japan. Sialidase (Artherobacter urifaciens) was produced by Novo Nordisk A/S and used as a 200 U/mL solution in 20 mM HEPES, 120 mM NaCl, 50% glycerol, pH 7.0. Support-bound neuraminidase (Clostridium perfringens) in agarose was obtained from Sigma (Cat. # N5254-10UN). The resin was washed extensively in MilliQ water before use. 10k-PSC (cytidine-5'-monophosphoryl-[5-(N-10kDa-methoxypolyoxyethylene-oxycarboxamido)glycylamido-3,5-dideoxy-D-glycero-n-galacto-2-nonulopyranosuronate], lot: 3605-A-R0-01-61-1) was obtained from Albany Molecular Research, Inc, Albany, NY 12203, USA. 10k-PSC was produced as described in WO03031464. The tris buffer used in the following examples contained 50 mM tris-HCl, 150 mM NaCl, pH 7.6.

SDS-PAGE gel electrophoresis was performed on NuPage precast gels (NuPage 7% Tris-Acetate Gel, 15 wells, Invitrogen cat. No. EA03555BOX) using NuPage Tris-Acetate SDS running buffer (Invitrogen cat. No. # LA0041). Samples (0.15 mg/mL) were spiked with NuPage Sample Reducing Agent (Invitrogen cat. No. NP0004) and LDS Sample Buffer (Invitrogen cat. No. NP0008) and heated at 70 °C for 10 min before gel loading. HiMark HMW (Invitrogen cat. No. LC6060) was used as a standard. Electrophoresis was performed at 150 V, 120 mA, and 25 W for 60 min. Gels were stained with SimplyBlue SafeStain (Invitrogen cat. no. LC5688) according to Invitrogen protocols. Coomassie-stained gels were analyzed by densitometry using GelQuant software and methods previously described (Rehbein and Schwalbe 2015) (FIG 34).

HPLC analysis of the glycoPEGylated EPO-Myc-His6 conjugates of the invention was performed as follows. Samples were analyzed in non-reduced form. A Zorbax 300SB-C3 column (4.6 x 50 mm; 3.5 um Agilent, Cat. No.: 865973-909) was used. The column was operated at 30 °C. 2.5 ug of sample was injected and the column was eluted with a water (A) - acetonitrile (B) solvent system containing 0.1% trifluoroacetic acid. The gradient program was as follows: 0 min (25% B); 4 min (25% B); 14 min (46% B); 35 min (52% B); 40 min (90% B); 40.1 min (25% B).

Enzymatic glycoPEGylation of EPO-Myc-His6 variants was carried out either by a one-step method using sialidase in combination with ST3Gal-III, or by a sequential method in which the EPO-Myc-His6 variant was first reacted with agarose-neuraminidase gel for 3 h at 25 °C, and after transferring the supernatant to a fresh vial it was reacted with ST3Gal-III reagent and PSC. The one-pot method is based on the fact that sialidase from Artherobacter urifaciens cannot desialylate PEGylated sialosides. The use of recombinant glycoproteins with single-antennary N-glycans without sialic acid protection as invented eliminates the need for prior desialylation.

GlycoPEGylation of Recombinant Human Erythropoietin: Recombinant wild-type EPO-Myc-His6 produced in CHO cell lines was dissolved in 100 mM Tris, 150 mM NaCl, 0.005% NaN3 (pH 7.2) at a concentration of 1.17 mg/mL. 10k-PSC reagent (2.0 mg) was dissolved in 100 uL of 50 mM Tris-HCl, 150 mM NaCl (pH 7.6) at a concentration of 20 mg/mL. Both ST3Gal-III and AUS were dissolved in 20 mM HEPES, 120 mM NaCl, 50% glycerol, pH 7.0 at a volumetric concentration of 1.35 U/mL and 200 U/mL respectively. The reagents were mixed according to the following scheme:

The reactions were incubated for 18 h at 32°C. Samples were then analyzed by HPLC and SDS-PAGE (FIG. 33, lanes 5-9).

GlycoPEGylation of Erythropoietin Produced in Mgat2/4A/4B/5 Knockout CHO Cells The recombinant wtEPO-Myc-His6 glycosylation variant produced in the Mgat2/4a/4b/5 knockout CHO cell line was dissolved in 100 mM Tris, 150 mM NaCl, pH 7.2 at a concentration of 0.585 mg/mL. 10k-PSC reagent (2.0 mg) was dissolved in 100 uL of 50 mM Tris-HCl, 150 mM NaCl, pH 7.6 at a concentration of 20 mg/mL. Both ST3Gal-III and AUS were dissolved in 20 mM HEPES, 120 mM NaCl, and 50% glycerol, pH 7.0, at volumetric concentrations of 1.35 U/mL and 200 U/mL, respectively. The reagents were mixed according to the following scheme:

The reactions were incubated for 18 h at 32°C. The samples were then analyzed by HPLC and SDS-PAGE (FIG. 22, lanes 11-15).

The two-step glycoPEGylation reaction of recombinant human erythropoietin-EPO-Myc-His6 (272 ug) in 100 mM Tris, 150 mM NaCl, 0.005% NaN3, pH 7.2 (0.214 mg/mL) was treated with pre-washed immobilized sialidase (150 uL suspension, 150 mU) for 3 h at room temperature. The resin was then removed by filtration. The filtrate was analyzed by LC-MS to confirm the formation of asialo EPO-Myc-His6 as a single-antenna N-glycan (calculated mass 25464.97 Da, analysis 25465.30 Da).

10k-PSC reagent dissolved in 100 uL of 50 mM Tris-HCl, 150 mM NaCl, pH 7.6, was then added to the filtrate. Both ST3Gal-III and AUS were dissolved in 20 mM HEPES, 120 mM NaCl, and 50% glycerol, pH 7.0, at volumetric concentrations of 1.35 U/mL and 200 U/mL, respectively. The reagents were mixed according to the following scheme:

The reactions were incubated for 20 h at 32°C. Samples were then analyzed by HPLC and SDS-PAGE.

More tables

Table 4. ZFN design and sequence analysis of glycoengineered CHO clones.

Clone Insert Alignment<k>

withdraw

KO: mgat4AZFN TTCTGAGTTGAATGCCATTGTCCAACAgtt----mgat4A-WTtcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

KO: mgat4A/4B

mgat4A-WTZFN TTCTGAGTTGAATGCCATTGTCCAACAgtt----tcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

mgat4B-WTZFN GCCCTCCAGCAGCCCTCTqaggaCTGGATGATCCTGGAGTTmgat4B-alle 1-7bp GCCCTCCAGCAGCCCCTCTg------GAT GAT CCT GGAGTTKO:mgat5

mgat5-WTZFN GGGGGATGATGCTTCTGCACTTCACCATCCAg----cagcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle 1+14bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcaTGAATTCTAGATGAgcgG

ACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle2-1 bp GGGGGATGATGCTTCTGCACTTCACCATCCAg----agcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle3+61bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcag----<( 61 b p ) —>cgGACTCAGCCTGAGAGCAGCTCCATGTT

KO: mgat4A/4B/5

mgat4A-WTZFN TTCTGAGTTGAATGCCATTGTCCAACAgtt----tcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

mgat4B-WTZFN GCCCTCCAGCAGCCCTCTqaggaCTGGATGATCCTGGAGTTmgat4B-alle 1-7bp GCCCTCCAGCAGCCCCTCTg-------GATGATCCTGGAGTTmgat5-WTZFN GGGGGATGATGCTTCTGCACTTCACCATCCAg---cagcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle 1+4bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcagcCAGCgGACTCAGCCTG

AGAGCAGCTCCATGTT

mgat5-alle2-4bp GGGGGATGATGCTTCTGCACTTCACCATCCAg----gGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle3-16bp GGGGGATGATGCTTCTGCACTTC------

CTCAGCCTGAGAGCAGCTCCATGTT

KO:B3gnt1

B3gnt1-WTZFN TGCAGCTGCTCTACCTGTC-----------gctgcTCTCCGGACTGCACGB3gnt1-alle1+11 bp TGCAGCTGCTCTACCTGTCgcagcTGCTCTACCTGTCTCCCGGA

CTGCACG

B3gnt1-alle2-5bp TGCAGCTGCTCTACCTGTC---------------- TCTCCGGACTGCACG

KO:B3gnt2

B3gnt2-WTZFN TTCAGCCCTTCCcgggcGTACTGGAACAGAGAGCAB3gnt2-alle 1-1bp TTCAGCCCTTCC-gggcGTACTGGAACAGAGAGCA

B3gnt2-alle2-4bp TTCAGCCCTTCCcg----T ACT GGAACAGAGAGCA

KO:B4gal tl

B4gal tl-WTZFN T GCAT CCGGTCCT ACAGCqccagc----AACT GGACT AT GGT AB4gal tl-alle1+4bp TGCATCCGGTCCTACAGCgccagcCAGCAACTGGACTATGGTA

KO:B4galt2

B4galt2-WTZFN CAGCCCCGCCACTTT qcc-----------atcGCCATGGACAAGTTTGGCTB4galt2-alle 1+73bp CAGCCCCGCCACTTTgcc--(+73bp)--atcGCCATGGACAAGTTTGGCT

KO:B4galt3

B4galt3-WTZFN CTAGCCCTCAAGTCAGGAtgt— tqCGGAGGCTGCTGGAGAGGB4galt3-alle 1+5bp CTAGCCCTCAAGTCAGGAtgtCGTGTtgCGGAGGCTGCTGGAGAGGB4galt3-alle2+2bp CTAGCCCTCAAGTCAGGAtgt---tgCCCGGAGGCTGCTGGAGAGG

KO:B4galt4

B4galt4-WTZFN AACT GGGACT GCTTT at----attcCACGAT GTGGACCTGGT GB4galt4-alle 1+1bp AACT GGGACT GCTTT at---T attcCACGAT GTGGACCT GGTGB4galt4-alle2+4bp AACT GGGACT GCTTT atattT ATT cCACGAT GTGGACCT GGTGKO:B4galt1/2

B4gal tl-WTZFN T GCAT CCGGTCCT ACAGCaccaac----AACT GGACT AT GGT AB4gal tl-alle1+4bp TGCATCCGGTCCTACAGCgccagcCAGCAACTGGACTATGGTAB4galt2-WTZFN CAGCCCCGCCACTTTaccatcGCCATGGACAAGTTTGGCTB4galt2- alle 1-8bp CAGCCCCGCCAC-------cGCCAT GGACAAGTTTT GGCT

B4galt2- alle2-14bp CAGCCC------------- cGCCAT GGACAAGTTT GGCT

KO:st3Gal3

st3Gal3-WTZFN CTCT CT CTTT GT CCTTGCtaacttCAAAT GGCAGGACTTCAAGst3Gal3- a lle l-5bp CTCT CT CTTT GT CCTTGCt-----CAAAT GGCAGGACTT CAAGst3Gal3- alle2-1bp CTCT CT CTTT GT CCTTGCtggct-CAAAT GGCAGGACTT CAAG

KO:st3Gal4

st3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCattat----aTTGTGGTGGGGAATGGGCst3Gal4-alle 1+4bp GGCAGCCTCCAGTGTCGTCgttgtTTGTgTTGTGGTGGGGAATGGGCst3Gal4- alle2-4bp GGCAGCCTCCAGTGTCGTCg--------gTTGTGGTGGGGAATGGGC

KO:st3Gal6

st3Gal6-WTZFN CGGT ACCT CT GATTTT GCTttgccCT AT GGACAAGGCC

st3Gal6- alle1-4bp CGGT ACCT CT GATTTT GCTt----CT AT GGGACAAGGCC

st3Gal6- alle2-22bp CGGTACCTCTGA------------AGGCC

KO:st3Gal3/4

st3Gal3-WTZFN CTCT CT CTTT GT CCTTGCtggcttCAAAT GGCAGGACTTCAAGst3Gal3- alle1-5bp CTCT CT CTTT GT CCTTGCt-----CAAAT GGCAGGACTT CAAGst3Gal3- alle2-1bp CTCT CT CTTT GT CCTTGCtggct-CAAAT GGCAGGACTT CAAGst3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCgttgt— gTTGTGGTGGGGAATGGGCst3Gal4-alle 1-4bp GGCAGCCTCCAGTGTCGTCg--------gTTGTGGTGGGGAATGGGCst3Gal4-alle2+5bp GGCAGCCTCCAGTGTCGTCgACACGttgtgTTGTGGTGGGGAATGGGC

KO:st3Gal4/6

st3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCattgt----aTTGTGGTGGGGAATGGGCst3Gal4-alle 1+4bp GGCAGCCTCCAGTGTCGTCgttgtTTGTgTTGTGGTGGGGAATGGGCst3Gal4- alle2-4bp GGCAGCCTCCAGTGTCGTCg--------gTTGTGGTGGGGAATGGGCst3Gal6-WTZFN CGGT ACCT CT GATTTT GCT-ttgccCT AT GGGACAAGGCC

st3Gal6-alle 1+1bp CGGT ACCT CT GATTTT GCTTttgccCT AT GGGACAAGGCC

st3Gal6-alle2-7bp CGGT ACCT CT GATTTT G------- CT AT GGGACAAGGCC

KO: B3gnt2/mgat4A/4B/5

mgat4A-WTZFN TTCTGAGTTGAATGCCATTGTCCAACAgtt----tcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

mgat4B-WTZFN GCCCTCCAGCAGCCCCTCTaaaaaCTGGATGATCCTGGAGTTmgat4B-alle 1-7bp GCCCTCCAGCAGCCCCTCTg-------G ATG ATCCTGG AGTTmgat5-WTZFN GGGGGATGATGCTTCTGCACTTCACCATCCAg---cagcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle 1+4bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcagcCAGCgGACTCAGCCTG

AGAGCAGCTCCATGTT

mgat5-alle2-4bp GGGGGATGATGCTTCTGCACTTCACCATCCAg----gGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5- alle3-16bp GGGGGATGATGCTTCTGCACTTC------

CTCAGCCTGAGAGCAGCTCCATGTT

B3gnt2-WTZFN TTCAGCCCTCC--caaacGTACTGGAACAGAGAGCA

B3gnt2-alle 1+2bp TTCAGCCCTTCCCCcgggcGTACTGGAACAGAGAGCA

B3gnt2-alle2+1bp TT CAGCCTTCCC-cgggcGT ACT GGAACAGAGAGCAKO:st3Gal4l6lmgat4AI4BI5

mgat4A-WTZFN TTCTGAGTTGAATGCCATTGTCCAACAgtt----tcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

mgat4B-WTZFN GCCCTCCAGCAGCCCCTCTaaaaaCTGGATGATCCTGGAGTTmgat4B- alle1-7bp GCCCTCCAGCAGCCCCTCTg------ GAT G AT CCTGG AGTTmgat5-WTZFN GGGGGATGATGCTTCTGCACTTCACCATCCAg---cagcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle 1+4bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcagcCAGCgGACTCAGCCTG

AGAGCAGCTCCATGTT

mgat5-alle2-4bp GGGGGATGATGCTTCTGCACTTCACCATCCAg----gGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle3-16bp GGGGGATGATGCTTCTGCACTTC------

CTCAGCCTGAGAGCAGCTCCATGTT

st3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCattat----aTTGTGGTGGGGAATGGGCst3Gal4-alle 1-5bp GGCAGCCTCCAGTGTCGTCgttgt-----aGTGGGGAATGGGCst3Gal4-alle2+4bp GGCAGCCTCCAGTGTCGTCaTTGTttataTTGTGGTGGGGAATGGGCst3Gal6-WTZFN CGGT ACCTCT GATTTTGCTttaccCT ATGGACAAGGCC

st3Gal6- alle1-4bp CGGTACCTCTGATTTTGCTt----CTATGGGACAAGGCCst3Gal6- alle2-2bp CGGTACCTCTGATTTTGCTtta--CTATGGGACAAGGCC

KI:ST6GAL 1IKO.:st3gal4I6

st3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCattat----aTTGTGGTGGGGAATGGGCst3Gal4-alle 1+4bp GGCAGCCTCCAGTGTCGTCattatTTGTaTTGTGGTGGGGAATGGGCst3Gal4- alle2-4bp GGCAGCCTCCAGTGTCGTCa------- aTT GT GGT GGGGAAT GGGCst3Gal6-WTZFN CGGT ACCTCT GATTTTGCT-ttaccCT AT GGGACAAGGCCst3Gal6-alle 1+1bp CGGTACCTCTGATTTTGCTTttaccCTATGGGACAAGGCCst3Gal6- alle2-7bp CGGT ACCTCT GATTTTG------- CTAGGGACAAGGCC

KI.ST6GAL11 KO:st3gal4/ 61 mgat4AI 4BI5

mgat4A-WTZFN TTCTGAGTTGAATGCCATTGTCCAACAgtttcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

mgat4B-WTZFN GCCCTCCAGCAGCCCCTCTaaaaaCTGGATGATCCTGGAGTTmgat4B- alle1-7bp GCCCTCCAGCAGCCCCTCTg------ GAT GAT CCTGG AGTTmgat5-WTZFN GGGGGATGATGCTTCTGCACTTCACCATCCAg---cagcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle 1+4bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcagcCAGCgGACTCAGCCTG

AGAGCAGCTCCATGTT

mgat5-alle2-4bp GGGGGATGATGCTTCTGCACTTCACCATCCAg----gGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle3-16bp GGGGGATGATGCTTCTGCACTTC------

CTCAGCCTGAGAGCAGCTCCATGTT

st3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCattat----aTTGTGGTGGGGAATGGGCst3Gal4-alle 1-5bp GGCAGCCTCCAGTGTCGTCattat-----aGTGGGGAATGGGCst3Gal4-alle2+4bp GGCAGCCTCCAGTGTCGTCaTTGTttataTTGTGGTGGGGAATGGGCst3Gal6-WTZFN CGGT ACCTCT GATTTTGCTttaccCT ATGGACAAGGCC

st3Gal6-alle 1-4bp CGGTACCTCTGATTTTGCTt----CTATGGGACAAGGCC

st3Gal6-alle2-2bp CGGTACCTCTGATTTTGCTtta--CTATGGGACAAGGCCKO:mgat2

mgat2-WTTALEN TCCTTTGTCGCCCATTGCTgctccagaggacgaagCCGCAGGCGGCCACCAC GA

mgat2-alle 1-4bp TCCTTTGTCGCCCATTGCTgctcc----gacgaagccgcaggcggccaccacga

KO: mgat2/stgal4/6

mgat2-WTTALEN TCCTTTGTCGCCCATTGCTgctccagaggacgaagCCGCAGGCGGCCACCAC GA

mgat2-alle 1-4bp TCCTTTGTCGCCCATTGCTgctccag---cgaagCCGCAGGCGGCCACCACGA

st3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCattat—-gTTGTGGTGGGGAATGGGCst3Gal4-alle 1+4bp GGCAGCCTCCAGTGTCGTCgttgtTTGTgTTGTGGTGGGGAATGGGCst3Gal4-alle2-4bp GGCAGCCTCCAGTGTCGTCg--------gTTGTGGTGGGGAATGGGCst3Gal6-WTZFN CGGT ACCT CT GATTTT GCT-ttgccCT AT GGGACAAGGCCst3Gal6-alle 1+1 bp CGGT ACCT CT GATTTT GCTTttgccCT AT GGGACAAGGCCst3Gal 6- alle2-7bp CGGT ACCT CT GATTTT G-------CT AT GGGACAAGGCC

KO: mgat2/mgat4A/4B/5

mgat2-WTTALEN TCCTTTGTCGCCCATTGCTgctccagaggacgaagCCGCAGGCGGCCACCAC GA

mgat2-alle 1-4bp TCCTTTGTCGCCCATTGCTgctccag---cgaagCCGCAGGCGGCCACCACGA

mgat4A-WTZFN TTCTGAGTTGAATGCCATTGTCCAACAgtt----tcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

mgat4B-WTZFN GCCCTCCAGCAGCCCTCTgaggaCTGGATGATCCTGGAGTTmgat4B-alle 1-7bp GCCCTCCAGCAGCCCCTCTg-------GATGATCCTGGAGTTmgat5-WTZFN GGGGGATGATGCTTCTGCACTTCACCATCCAg---cagcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle 1+4bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcagcCAGCgGACTCAGCCTG

AGAGCAGCTCCATGTT

mgat5-alle2-4bp GGGGGATGATGCTTCTGCACTTCACCATCCAg----gGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle3-16bp GGGGGATGATGCTTCTGCACTTC------

CTCAGCCTGAGAGCAGCTCCATGTT

KO:mgat2/st3gal4/6/mgat4A/4B/5

mgat2-WTTALEN TCCTTTGTCGCCCATTGCTgctccagaggacgaagCCGCAGGCGGCCACCAC GA

mgat2- alle1-5bp TCCTTTGTCGCCCATTGCTgctcca----cgaaGCCGCAGGCGGCCACCACGA

mgat4A-WTZFN TTCTGAGTTGAATGCCATTGTCCAACAgtt----tcGCCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle 1-5bp TTCTGAGTTGAATGCCATTGTCCAACAg----

CCGTGCAGGAGCAGCAACTAACGGAAG

mgat4A-alle2+13bp TTCTGAGTTGAATGCCATTGTCCAACAgttGAATTCTAGATGAtcGCCGTGC

AGGAGCAGCAACTAACGGAAG

mgat4B-WTZFN GCCCT CCAGCAGCCCCT CTgaggaCT GGAT GATCCT GGAGTTmgat4B- a lle l-7bp GCCCTCCAGCAGCCCCTCTg-------GATGATCCTGGAGTTmgat5-WTZFN GGGGGATGATGCTTCTGCACTTCACCATCCAg---cagcgGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle 1+4bp GGGGGATGATGCTTCTGCACTTCACCATCCAgcagcCAGCgGACTCAGCCTG

AGAGCAGCTCCATGTT

mgat5-alle2-4bp GGGGGATGATGCTTCTGCACTTCACCATCCAg----gGACTCAGCCTGAGAGCAGCTCCATGTT

mgat5-alle3-16bp GGGGGATGATGCTTCTGCACTTC------

CTCAGCCTGAGAGCAGCTCCATGTT

st3Gal4-WTZFN GGCAGCCTCCAGTGTCGTCattat----gTTGTGGTGGGGAATGGGCst3Gal4-alle 1-5bp GGCAGCCTCCAGTGTCGTCgttgt-----gGTGGGGAATGGGCst3Gal4-alle2+4bp GGCAGCCTCCAGTGTCGTCgTTGTttgtgTTGTGGTGGGGAATGGGCst3Gal6-WTZFN CGGT ACCT CT GATTTT GCTttgccCT AT GGACAAGGCCst3Gal6- a lle l-4bp CGGT ACCT CT GATTTT GCTt----CT AT GGGACAAGGCCst3Gal6-alle2-2bp CGGT ACCT CT GATTTT GCTttg--CT AT GGGACAAGGCC

KO:futB

fut8-WTCRISPR CAAATACTTGATCCGTCCACAACCT-TGGCTGGAAAGGGAAfut8 - alle1-1 bp CAAATACTTGATCCGTCCACAACC--TGGCTGGAAAGGGAAfut8-alle2+1bp CAAAT ACTT GAT CCGTCCACAACCTTTGGCT GGAAAGGGAA

KO:futB/ B4galt1

fut8-WTCRISPR CAAATACTTGATCCGTCCACAACCT-TGGCTGGAAAGGGAAfut8- alle1-4bp CAAAT ACTT GAT CCGTCCACAA-----GGCT GGAAAGGGAAfut8-alle2+1bp CAAAT ACTT GAT CCGTCCACAACCTTTGGCT GGAAAGGGAAB4galtl-WTZFN TGCATCCGGTCCTACAGCgccaqc----AACTGGACTATGGTAB4gal tl- alle1+4bp T GCAT CCGGT CCT ACAGCgccagcCAGCAACT GGACT AT GGT A

KO:mgat3

mgat3-WTZFN TTCCTGGACCACTTCCCAcccggt----------GGCCGGCAGGATGGCmgat3- alle1-21bp TTCCT GGACCACT------------------------------- ATGGC

mgat3-alle2+293bp TTCCT GGACCACT GAT ACcc--+293bp--cggtGGCCGGCAGGAT GGC

KO:mgat4C

mgat4C-WTZFN AT ACTT CAGACT AT tatgtAATGCT CGAAGAT GAT GTT

mgat4C- alle1-13bp AT ACTT CAGACT------------ CGAAGAT GAT GTT

mgat4C-alle2-8bp AT ACTT CAGACT AT------- GCTCGAAGAT GAT GTT

KO:mgat5B

mgat5B-WTZFN CGT GGCGCCCT CCGCAAGatgagtGACCT GCTGGAGCT Gmgat5B-alle 1-7bp CAGCTCCAGCAGGT-------CTTGCGGAGGGCGCCACGmgat5B-alle2-5bp CAGCTCCAGCAGGT— ATCTTGCGGAGGGCGCCACG

KO:B3gnt8

B3gnt8-WTTALEN TGGTCCAGAGATAGCTAATgaagcttctagggtgGAGAAGCTGGGGCTGCTG A

B3gnt8-alle 1-17bp TGGTCCAGAGG-----------------

AGGGTGGAGAAGCTGGGGGCTGCTGA

KO:mgat1

mgat1-WTZFN AACAAGTT CAAGTT CccagcaGCT GT GGT AGT GGAGGACmgat1-alle1-2bp AACAAGTT CAAGTT CCc--gcaGCT GTGGT AGTGGAGGAC

Target region of the gene underlined.

Table 5

tcgcgcgtttcggtgatgacggtgaaaacctctgacacatgcagctcccggagacggtcacagcttgtctgta agcggatgccgggagcagacaagcccgtcagggcgcgtcagcgggtgttggcgggtgtcggggctggcttaac tatgcggcatcagagcagattgtactgagagtgcaccatatgcggtgtgaaataccgcacagatgcgtaagga gaaaataccgcatcaggcgccattcgccattcaggctgcgcaactgttgggaagggcgatcggtgcgggcctc ttcgctattacgccagctggcgaaagggggatgtgctgcaaggcgattaagttgggtaacgccagggttttcc cagtcacgacgttgtaaaacgacggccagtgaattcgagctcggtaccaagcttggttgcatgctgtccggag tctcagcgttataccagaagtgacctgggtcggggaagactatagtgtcacctaaatctctctagagcccttcat taggcgcgccaatcccattgcaaattctacaaaaggagtgtttcccaactgctctatcaagaggaatgttgca cactgtgacctgaatgcaaacatcacacgcgccagcagagaggaagaagagaggcttccctgaccgggaatcg aacccgggccgcggcggtgagagcgccgaatcctaaccactagaccaccagggagcacgcgccaaagctcaat gagctataattatccccttggaaaacctacaaaaacagtgtttcaaaactgctctgtgaaaagggacctttgc tagcacgcggcgccaggcaaaacgtgggcacgctgcgttggccgggaatcgaacccgggtcaactgcttggaa ggcagctatgctcaccactataccaccaacgcgcacacgcgccagcagattctacgggaagagtgtttcaaaa ctgctctatcaagagaaatgttccaccttgtgtgtggaatgcagccatcacacgcgtccatgaaagggcttaa ttaagatatcgtttaaacgtcgacctgcagaggccggcggataactagctgatcgcggaatcctgtccctagg ccacccactgtggggtgcccttcattaggcgcgccaatcccattgcaaattctacaaaaggagtgtttcccaa ctgctctatcaagaggaatgttgcacactgtgacctgaatgcaaacatcacacgcgccagcagagaggaagaa gagaggcttccctgaccgggaatcgaacccgggccgcggcggtgagagcgccgaatcctaaccactagaccac cagggagcacgcgccaaagctcaatgagctataattatccccttggaaaacctacaaaaacagtgtttcaaaa ctgctctgtgaaaagggacctttgctagcacgcggcgccaggcaaaacgtgggcacgctgcgttggccgggaa tcgaacccgggtcaactgcttggaaggcagctatgctcaccactataccaccaacgcgcacacgcgccagcag attctacgggaagagtgtttcaaaactgctctatcaagagaaatgttccaccttgtgtgtggaatgcagccat cacacgcgtccatgaaagggcggttgcatgctgtccggagtctcagcgttataccagaagtgacctgggtcgg ggaagaaaagcttcatggtcatagctgtttcctgtgtgaaattgttatccgctcacaattccacacaacatac gagccggaagcataaagtgtaaagcctggggtgcctaatgagtgagctaactcacattaattgcgttgcgctc actgcccgctttccagtcgggaaacctgtcgtgccagctgcattaatgaatcggccaacgcgcggggagaggc ggtttgcgtattgggcgctcttccgcttcctcgctcactgactcgctgcgctcggtcgttcggctgcggcgag cggtatcagctcactcaaaaggcggtaatacggttatccacagaatcaggggataacgcaggaaagaacatgtg agcaaaaggccagcaaaaggccaggaaccgtaaaaaggccgcgttgctggcgtttttccataggctccgcccc cctgacgagcatcacaaaaatcgacgctcaagtcagaggtggcgaaacccgacaggactataaagataccagg cgtttccccctggaagctccctcgtgcgctctcctgttccgaccctgccgcttaccggatacctgtccgcctt tctcccttcgggaagcgtggcgctttctcatagctcacgctgtaggtatctcagttcggtgtaggtcgttcgc tccaagctgggctgtgtgcacgaaccccccgttcagcccgaccgctgcgccttatccggtaactatcgtcttg agtccaacccggtaagacacgacttatcgccactggcagcagccactggtaacaggattagcagagcgaggta tgtaggcggtgctacagagttcttgaagtggtggcctaactacggctacactagaagaacagtatttggtatc tgcgctctgctgaagccagttaccttcggaaaaagagttggtagctcttgatccggcaaacaaaccaccgctg gtagcggtggtttttttgtttgcaagcagcagattacgcgcagaaaaaaaggatctcaagaagatcctttgat cttttctacggggtctgacgctcagtggaacgaaaactcacgttaagggattttggtcatgagattatcaaaa aggatcttcacctagatccttttaaattaaaaatgaagttttaaatcaatctaaagtatatatgagtaaactt ggtctgacagttagaaaaactcatcgagcatcaaatgaaactgcaatttattcatatcaggattatcaatacc atatttttgaaaaagccgtttctgtaatgaaggagaaaactcaccgaggcagttccataggatggcaagatcc tggtatcggtctgcgattccgactcgtccaacatcaatacaacctattaatttcccctcgtcaaaaataaggt tatcaagtgagaaatcaccatgagtgacgactgaatccggtgagaatggcaaaagtttatgcatttctttcca gacttgttcaacaggccagccattacgctcgtcatcaaaatcactcgcatcaaccaaaccgttattcattcgt gattgcgcctgagcgagacgaaatacgcgatcgctgttaaaaaggacaattacaaacaggaatcgaatgcaacc ggcgcaggaacactgccagcgcatcaacaatattttcacctgaatcaggatattcttctaatacctggaatgc tgttttcccagggatcgcagtggtgagtaaccatgcatcatcaggagtacggataaaatgcttgatggtcgga agaggcataaattccgtcagccagtttagtctgaccatctcatctgtaacatcattggcaacgctacctttgc catgtttcagaaacaactctggcgcatcgggcttcccatacaatcgatagattgtcgcacctgattgcccgac attatcgcgagcccatttatacccatataaatcagcatccatgttggaatttaatcgcggcctagagcaagac gtttcccgttgaatatggctcatactcttcctttttcaatattattgaagcatttatcagggttattgtctca tgagcggatacatatttgaatgtatttagaaaaataaacaaataggggttccgcgcacatttccccgaaaagt gccacctgacgtctaagaaaccattattatcatgacattaacctataaaaataggcgtatcacgaggcccttt cgtc

Claims (13)

1. A mammalian cell in which mgat4A, mgat4B, and mgat5 are deleted in the cell. 2. The cell according to claim 1, wherein the genes have been inactivated using zinc fingers, TALENS, or CRISPR/Cas 9. 3. The cell according to any of claims 1-2, with the elimination of mgat4A, mgat4B, and mgat4C. 4. The cell according to any of claims 1-3, comprising the removal of mgat5 and mgat5B. 5. The cell according to one of claims 1-4, wherein FUT8 has been removed. 6. The cell according to one of claims 1-5, wherein B3gnt2 has been removed. 7. The cell according to any of claims 1-6, wherein mgat2, mgat4A, mgat4B and mgat5 have been deleted. 8. The cell according to any of claims 1-7, wherein ST3gal3, ST3gal4 and ST3gal6 have been removed. 9. The cell according to one of claims 1-8, wherein FUT8.ST6Gal1 has been activated. 10. The cell according to any one of claims 1-9, wherein the cell is selected from the group consisting of CHO, NS0, SP2/0, YB2/0, CHO-K1, CHO-DXB11, CHO-DG44, CHO-S, HEK293, HUVEC, HKB, PER-C6, NS0, or derivatives of any of these cells. 11. The cell according to any of claims 1-10, wherein the cell is a CHO cell. 12. The cell according to any one of claims 1-11, further encodes an exogenous protein of interest, which is a polypeptide selected from the group consisting of EPO, an IgG antibody, a protein involved in hemostasis, or a coagulation factor. 13. A method for producing a recombinant protein with a more homogeneous glycosylation pattern, the method comprising: - expressing the protein in a mammalian cell according to any of the preceding claims.